City large-area power failure scenario simulation and deduction method and system based on digital twinning
By constructing a digital twin model to simulate large-scale urban power outage scenarios, the problem of lack of pre-event dynamic simulation in existing technologies has been solved. This enables dynamic simulation and multi-system coupled simulation of urban power outage disaster scenarios, improving the scientific nature and predictive ability of emergency decision-making.
Patent Information
- Application Number
- CN202611228810.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-13
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies for analyzing and managing the risks of large-scale power outages in cities lack the ability to dynamically predict major power outage disasters in advance. They are unable to identify key decision-making time points in the evolution of disasters and quantify the effectiveness of emergency strategies, resulting in a long-term reactive response.
A digital twin-based method for simulating large-scale power outages in cities constructs digital twin models of key infrastructure such as power grids, transportation, and water supply. It then combines multi-source operational data to conduct simulations, identify key time windows, and visualize them, enabling quantitative comparison of multiple strategies and interactive three-dimensional visualization.
It significantly enhances the realism and dynamism of urban power outage disaster scenario construction, provides scientific pre-verification methods, and forms a complete closed-loop decision support capability encompassing simulation, strategy verification, contingency plan generation, and continuous evolution, thus addressing the need for proactive decision-making throughout the entire process.
Smart Images

Figure CN123117940A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of urban safety, emergency management and digital twin technology, and more specifically, to a method and system for simulating and extrapolating large-scale power outage scenarios in cities based on digital twins. Background Technology
[0002] Existing technologies for risk analysis and emergency management of large-scale urban power outages are deficient in that they lack the ability to dynamically simulate power outage disasters in advance. They cannot systematically simulate the chain evolution of power outage disasters among multiple urban systems before or in the early stages of a disaster, cannot identify key decision-making time points in the disaster evolution, and cannot quantitatively predict the effects of different emergency strategies. This leads to a long-term technical dilemma of reactive response to disasters, making it difficult to meet the needs of proactive decision-making throughout the entire process in the context of large-scale urban power outage disasters. Summary of the Invention
[0003] In view of this, the present invention proposes a method and system for simulating and extrapolating large-scale urban power outage scenarios based on digital twins, in order to solve the problems existing in the prior art.
[0004] To achieve the above objectives, this invention proposes a method for simulating and extrapolating large-scale urban power outage scenarios based on digital twins, comprising: Acquire multi-source operational data of the city; Construct a digital twin model of infrastructure, wherein the digital twin model of infrastructure includes a power grid sub-model, a transportation sub-model and a water supply sub-model, wherein the power grid sub-model, the transportation sub-model and the water supply sub-model are coupled and related. Obtain the triggering parameters for power outage scenarios; based on the multi-source operational data and the digital twin model of the infrastructure, simulate and extrapolate the triggering parameters for power outage scenarios to obtain the disaster simulation and extrapolation results and key time windows; The results of disaster simulation and key time windows are visualized.
[0005] Optionally, the multi-source operational data includes power grid operational data, spatial data, meteorological data, traffic operational data, base station operational data, and water and gas supply operational data; Optionally, in the infrastructure digital twin model, the power grid sub-model uses generators, buses, and loads as node objects and lines as edge objects, and generates a topology structure based on the node objects and edge objects, and sets a state estimator based on the topology results; the traffic sub-model is constructed by dividing the corresponding cells according to road length and traffic capacity; and the water supply sub-model is constructed based on the network composed of water supply pipeline nodes and pipe segments.
[0006] Optionally, the power grid sub-model, transportation sub-model, and water supply sub-model are coupled and correlated according to a coupling correlation matrix; wherein the coupling correlation matrix includes coupling elements between different systems; wherein the different systems include power grid, transportation, communication, water supply, and gas systems; and the coupling elements include strength coefficient, propagation probability, and time delay.
[0007] Optionally, the process of simulating and extrapolating the disaster triggering parameters for power outage scenarios includes: Obtain the triggering parameters for power outage scenarios; based on the multi-source operational data and the infrastructure digital twin model. The multi-source operational data is used as the initial state of the infrastructure digital twin model; the parameters of the infrastructure digital twin model in the initial state are adjusted by triggering parameters; transient and steady-state fault simulations are performed using the adjusted infrastructure digital twin model; during the fault simulation process, secondary disaster chain evolution simulation, critical time window identification, and emergency resource allocation simulation are performed to obtain disaster simulation and deduction results and critical time windows.
[0008] Optionally, the process of simulating the evolution of the secondary disaster chain includes: Within each simulation time step, the traffic agent, communication agent, and water supply agent in the infrastructure digital twin model adjust their states. At the end of each time step, each agent writes its own state into the coupling correlation matrix and reads the states of other systems, and obtains the failure probability. The failure probability is calculated based on the failure contribution probability of the corresponding upstream path. Each agent adds the comprehensive failure probability to the state calculation of the next time step.
[0009] Optional, key time windows include: The timeframes for power outage triggering, initial cascading failure initiation, initial secondary disaster initiation, golden window cutoff, critical infrastructure mass failure, and recovery restart are determined. The initial secondary disaster initiation time is obtained by predicting the final power outage ratio. The golden window cutoff time is obtained by weighted summation of the failure states of the five active propagation systems in the coupling correlation matrix and threshold determination. The critical infrastructure mass failure time and recovery restart time are obtained by backfilling the time when the power grid begins to restore power supply in the optimal solution.
[0010] Optional, the emergency resource allocation simulation process includes: A resource database is constructed. When a resource allocation request is triggered, available resources are filtered from the resource database based on the triggered resource allocation request. Using real-time traffic conditions provided by the traffic digital twin model as input, a dynamic shortest path algorithm is used to calculate the recommended path and estimated travel time. The optimal resource combination is selected according to the estimated arrival time to form an allocation plan. The allocation plan is simulated to obtain the recovery effect, and the recovery effect is fed back to the state calculation of each infrastructure agent in real time.
[0011] On the other hand, the present invention provides a simulation and deduction system for large-scale urban power outage scenarios based on digital twins, for performing the above-described methods.
[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention significantly enhances the realism and dynamism of urban power outage disaster scenario construction through the comprehensive application of technologies such as multi-system coupled digital twin modeling, spatiotemporal dynamic evolution simulation of power outages, automatic identification of critical time windows, parallel extrapolation of multiple scenarios, and model self-evolution. It achieves a qualitative leap from static assessment to dynamic evolution, from single-system analysis to multi-system coupled extrapolation, and from experience-based decision-making to time-quantitative driven decision-making. On this basis, through multi-strategy quantitative comparison and three-dimensional visualization interaction, it provides a scientific pre-verification method for emergency decision-making. Furthermore, through the model self-evolution mechanism, the system accuracy is continuously improved, and the emergency plan can be dynamically optimized according to the extrapolation results. Ultimately, it forms a complete closed-loop decision support capability encompassing simulation extrapolation, strategy verification, plan generation, and continuous evolution, effectively solving the technical challenge that existing technologies cannot meet the needs of proactive decision-making throughout the entire process under urban large-scale power outage disaster scenarios. Attached Figure Description
[0013] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. In the drawings: Figure 1 This is a schematic diagram of the method flow in an embodiment of the present invention. Detailed Implementation
[0014] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0015] Current power outage risk analyses largely remain at the static assessment stage, lacking the system simulation capabilities to depict the dynamic diffusion process, critical time windows, and coupled impacts of power outage disasters in urban spaces, thus failing to meet the full-process decision-making needs under catastrophic scenarios. This invention constructs digital twin models of critical infrastructure such as urban power grids, transportation, communications, and water supply, using power outage events as triggers and combining real-time or preset disaster parameters to simulate the evolution of power outage events across different time scales and spatial ranges. By constructing a coupled digital twin model of urban infrastructure and integrating power grid operation, meteorological, hydrological, transportation, and communication data, it simulates the multi-dimensional spatiotemporal evolution of power outage events. It can generate spatial diffusion paths, impact ranges, and critical time windows under different power outage scenarios, enabling dynamic simulation and decision support for large-scale urban power outage catastrophic events. This invention enhances the realism, dynamism, and decision support capabilities of power outage catastrophic scenario construction.
[0016] like Figure 1 As shown, the present invention provides a method for simulating and extrapolating large-scale urban power outage scenarios based on digital twins, which includes the following: 1. Data acquisition.
[0017] First, phasor measurement units are deployed in each substation in the city to collect synchronous phasor data of voltage and current. At the same time, switch status, protection action signals and equipment operating parameters are obtained from the data acquisition and monitoring system. After time synchronization and marking at the station end, the two types of data are transmitted to the central data platform via the power data network.
[0018] Secondly, spatial data processing is involved. Road network layers, building outline layers, and key facility distribution layers are exported from the urban geographic information system. All layers are then converted into a universal transverse Mercator projection coordinate system, and a spatial grid index is established to accelerate subsequent spatial query operations.
[0019] The access to meteorological and hydrological data is achieved by establishing dedicated data lines with the meteorological bureau and the water resources bureau. The system receives minute-by-minute observation data from automatic weather stations and hourly forecast data from numerical weather prediction in real time. After receiving the data, the system performs format parsing and quality control, removes obvious outliers, and then maps the station data to the spatial location of the power grid equipment using the inverse distance weighted interpolation method.
[0020] The collection of infrastructure status data relies on an Internet of Things (IoT) sensor network that is spread throughout the city. Traffic signal controllers report their operating status and power supply mode every few seconds, communication base stations report the remaining battery power and backup generator status every minute, water supply network pressure monitoring points report water pressure values, and gas pressure regulating stations report power supply status and outlet pressure. Different types of data are transmitted using different message topics for classification and processing.
[0021] The generation of dynamic population distribution data involves integrating base station-level signaling data provided by mobile operators, location service data from mainstream map applications, and public transportation card swiping data. Population census data is used as the total baseline for calibration, and finally, a population distribution matrix with a grid resolution of 250 meters at hourly intervals is output.
[0022] After all data enters the central platform, it first enters the data buffer queue, where it undergoes format standardization, missing value imputation, and time alignment operations in sequence. After processing, the data is stored in the real-time database and the historical database for subsequent use. When various types of data enter the simulation engine, they do not need to be uniformly resampled to a single frequency. Instead, the simulation engine adaptively selects the nearest sampled value or linear interpolation method to complete the time alignment based on the current simulation stage.
[0023] 2. Digital twin model construction.
[0024] Based on the various data obtained in step one, sub-models for power grid, transportation, and water supply are established respectively.
[0025] The construction of the power grid sub-model involves exporting a common information model file of the power grid from the dispatch automation system, parsing the conductive equipment, connection nodes, and topological relationships, and converting it into a graph data structure. Generators, buses, and loads are treated as node objects, while transmission lines, distribution lines, and transformers are treated as edge objects. Each object includes its static parameters such as rated voltage, rated capacity, and resistance / reactance values. A state estimator is then built upon this static topology. This estimator takes measurements provided by the SCADA system as input and uses a weighted least squares method to solve for the optimal estimates of the state variables, namely the voltage amplitude and phase angle of each node, thus forming a complete profile of the power grid's operating state.
[0026] The state estimation result, after low-pass filtering and smoothing, serves as the initial steady-state operating point for subsequent dynamic simulations. The filtering stage only eliminates high-frequency jitter introduced by measurement noise, without altering the unbiasedness of the estimated value. After outputting the voltage amplitude and phase angle of each node from the state estimation, the initial power angle of each generator is inferred from the bus voltage and generator-injected power by solving the power balance equations of each generator. The initial speed is uniformly set to synchronous speed, thus forming a complete initial state for dynamic simulation. If the system is already in a non-steady state at the user-defined initial moment, such as during the transient process immediately after a fault, the power balance equations may not converge, leading to failure in inferring the power angle. In this case, the system automatically uses the default power flow calculation result as the initial value, or the user can directly input the initial power angle values of each generator in the scenario parameter configuration interface.
[0027] At this steady-state operating point, a dynamic model library is loaded, including the generator's rotor motion equations and excitation regulator model, the load's voltage static characteristic model, and the action logic models of various relay protections. These models are embedded in the form of numerical differential equations and are solved iteratively as the state changes during the simulation process.
[0028] This invention covers both transmission and distribution network levels. For transmission networks, the Newton-Raphson method is used to solve the power flow equations. For distribution networks, due to their radial structure and high resistance-to-reactance ratio, the forward-backward substitution method is used for power flow calculation. Fault propagation at the distribution network level requires additional consideration of the operating logic of automatic transfer switches (ATS) and the switching conditions for islanded operation of distributed generation. When the number of tie switches forming closed-loop operation due to ATS or islanded switching in the distribution network exceeds 5% of the total number of tie switches, the forward-backward substitution method may experience slower convergence or even divergence due to the presence of weak loops. The system automatically switches to the Newton-Raphson method for solving the problem. Before the switch, the final voltage distribution after convergence by the forward-backward substitution method is used as the initial iteration value for the Newton-Raphson method to ensure the numerical continuity of the calculation results before and after the switch.
[0029] The traffic sub-model is constructed starting from road network GIS data. Each road is divided into several cells of equal length according to its length and capacity. Each cell records its vehicle density and average speed at any time. The amount of vehicle exchange between cells is determined by the sending capacity of the upstream cell and the receiving capacity of the downstream cell. Signalized intersections are modeled as special cells with periodic capacity reduction, and the reduction coefficient is obtained by mapping the power supply status of the intersection.
[0030] The construction of the water supply sub-model involves abstracting the water supply network into a network composed of nodes and pipe segments. Each node records its elevation and head value, and each pipe segment records its length, diameter, and roughness coefficient. Based on this, a set of hydraulic balance equations is established, and the water pressure of each node and the flow rate of each pipe segment are obtained through iterative solution.
[0031] 3. Coupling of multiple infrastructures.
[0032] Based on the establishment of the subsystem models in step two, a description of the dependencies between the systems is constructed.
[0033] A five-row, five-column coupling correlation matrix is established, with rows and columns corresponding to the five systems: power grid, transportation, communication, water supply, and gas supply. The medical system only serves as a disaster-bearing entity, receiving failure propagation from other systems. It does not generate a reverse propagation effect on other systems, so it is removed from the active propagation nodes of the coupling matrix and treated as a separate disaster-bearing entity for assessment.
[0034] Each non-zero element in the matrix is a triplet data structure containing three fields: intensity coefficient, propagation probability, and time delay. The intensity coefficient is a floating-point number between zero and one, representing the proportion of the source system failure's impact on the target system's function. The propagation probability is also a floating-point number between zero and one, representing the conditional probability that the target system will fail given the source system failure. The time delay is a non-negative integer in minutes, representing the time difference between the occurrence of the source system failure and the beginning of observable failure in the target system. Taking the coupling relationship between the power grid and traffic as an example, the intensity coefficient is determined by statistically analyzing data on traffic light failures and reduced intersection capacity during historical power outages. The propagation probability is determined by statistically analyzing the proportion of traffic light failures after a power outage. The time delay is directly recorded as the delay time for the traffic lights to switch to backup battery power.
[0035] After constructing the coupling matrix, a circular dependency detection is performed, scanning all bidirectional dependency pairs. When a bidirectional dependency is detected between any two systems, the failure propagation probability in one direction is automatically multiplied by a reduction factor. The reduction direction prioritizes the physically weaker dependency, and the reduction factor defaults to 0.5, which can be adjusted by the user in the system configuration according to the actual physical characteristics. If a bidirectional dependency still exists after reduction, the reduction continues until unidirectional logic dominates, thereby eliminating the risk of logic oscillation. If the time delay τ of a dependency is greater than the total duration of the current simulation, the system provides a prompt but allows the user to confirm and continue execution. This dependency does not produce an observable failure propagation effect within this simulation cycle.
[0036] Once the digital twin model is built, it enters the running state. Real-time data streams continuously update the state variables in the model. At the same time, the model performs state verification by comparing the measured values at fixed intervals. When the cumulative deviation exceeds the threshold, incremental updates are triggered.
[0037] 4. Input parameters for triggering the catastrophe.
[0038] With the digital twin model ready in step three, the user sets the power outage triggering conditions for this simulation through the scenario parameter configuration interface. The system first displays the scenario parameter configuration interface, with a scenario type selection area at the top, providing six preset template buttons: typhoon, rainstorm and flood, freezing disaster, equipment failure, network attack, and load surge. After the user clicks on any template, the system automatically loads the default parameter combination for that type.
[0039] The central part of the interface is the parameter editing area. Users need to set the initial spatial coordinates of the fault one by one. The operation method is to click on the map to select the location of the substation or line tower, and the system will automatically read the equipment number and voltage level at that location; set the specific date and time of the fault occurrence, supporting the selection of the current time, historical time or future preset time; set the electrical characteristics of the fault, including short circuit type, transition resistance value and fault duration; set the protection action parameters, including fault clearing time and reclosing method; set the external environmental parameters, including meteorological conditions such as wind speed, precipitation intensity and temperature.
[0040] The bottom of the interface is the scenario saving area. After completing the parameter configuration, users can choose to save this scenario as a new template or overwrite an existing template. During saving, the system automatically extracts all currently configured parameter values, serializes them into a standard format, and stores them in the scenario library. The system also provides a historical case retrieval function. After the user enters disaster keywords in the search box, the system performs text matching and vector similarity calculation in the case library, returning several historical cases with the highest similarity and their complete parameter configurations. Users can directly load the parameters of any historical case, modify them, and then use them. After completing this step, the simulation environment is fully ready.
[0041] 5. Spatiotemporal evolution simulation of power outage.
[0042] This step is performed concurrently with step six. The cascading failure simulation is divided into two cascading stages: transient and steady state.
[0043] During the transient phase, the system injects a short-circuit fault at the user-defined fault location and type. It then calls upon an electromagnetic transient simulation program to calculate the transient changes in bus voltage and branch current over several cycles following the fault occurrence. Based on the protection settings, it determines which protection devices should operate, and after executing the protection actions, it identifies the equipment to be disconnected and the initial power outage area. During the transient phase, electromechanical transient simulation is performed to capture protection actions triggered by power angle instability or voltage collapse during the transient process.
[0044] The termination criterion for the transient phase is set as follows: when the power grid is still connected to the grid, the criterion is that the absolute value of the system frequency deviation is less than 0.05 Hz and the amplitude deviation of all bus voltages is less than 0.05 per unit, and this state is maintained for at least 0.5 seconds of simulation time. If the power grid has split into multiple islands that are not synchronized with the main grid, the criterion is changed to the frequency change rate within all islands being less than 0.1 Hz per second. After each electromechanical transient simulation step, the system performs a power grid connectivity test. If the power grid topology is detected to change from single connectivity to multiple connectivity, the termination criterion is immediately switched to the island mode criterion. The transient is considered to have ended when either condition is met.
[0045] After the transient process subsides, the system automatically enters the steady-state phase of the overload protection iterative loop. Each iteration of this loop first performs power flow calculations based on the current grid topology and operating status, checking the load rate of all operating lines and the voltage amplitude of all buses. If there are lines with a load rate exceeding 100%, the inverse-time overload protection action time for each overloaded line is calculated, and the line with the shortest action time is prioritized for disconnection. After each disconnection, the system immediately returns to the power flow calculation step to solve the problem again. If there are buses with voltages below the allowable lower limit, they are marked as voltage instability nodes, and voltage stability is restored through load shedding. Each iteration also records the geographical location and time of disconnection of the disconnected equipment.
[0046] During the dynamic tracking of power outage boundaries, the system maintains a set of polygons representing the power outage area. After each topology change, it recalculates the minimum convex hull of all outage equipment or generates the power outage area boundary through a rasterization method. The boundaries at different times are stored as time series data, and the power outage area and diffusion rate at each time are calculated accordingly.
[0047] The termination condition for a chain failure iteration is any one of the following three situations: no new equipment is cut off between two consecutive iterations and all operating parameters are within the allowable range; the power outage ratio exceeds the preset collapse threshold; the absolute value of the change in the power outage ratio between five consecutive iterations is less than 0.5 percent.
[0048] 6. Simulation of the evolution of secondary disaster chains.
[0049] This step is executed concurrently with step five. The system uses a fixed time step of one minute, which is consistent with the minute unit of the time delay τ in the coupling correlation matrix. Within each simulation time step, each infrastructure agent independently calculates its own state changes.
[0050] The traffic intelligence agent updates the capacity of each signalized intersection based on the power supply status of the intersection signals. If the signal fails completely, the capacity is reduced to 30% to 40% of the normal value. The traffic flow model is then used to calculate vehicle location updates and congestion index changes. The communication intelligence agent calculates the set of base stations still in operation based on the remaining battery power of each base station. When a base station runs out of power, it is marked as offline, and the expansion range of the signal coverage blind spot is calculated. The water supply intelligence agent re-solves the hydraulic balance equations based on the power supply status of the water plant and pumping stations, outputs the water pressure values for each node, and marks nodes with water pressure below the minimum service head and their surrounding areas as water outage areas.
[0051] At the end of each time step, each agent writes its current state into the corresponding row of the coupling correlation matrix, and simultaneously reads the latest states of other systems from the matrix, calculating the correction amount for the impact of external failures on the functionality of its own system based on dependencies. For the failure state of the target system S at time t, the system traverses all its upstream systems S.i Read the upstream system at time (t-τ) i The historical failure state value of the target system is multiplied by the failure propagation probability α from the upstream system to the target system. i Then multiply by the upstream system's dependence strength coefficient F on the target system. i After obtaining the failure contribution probability of a single upstream path, the combined failure probability is calculated by taking the contributions of all upstream paths as follows: The combined failure probability is equal to one minus the product of the "non-failure probabilities" of all upstream paths, where the non-failure probability of each upstream path is equal to one minus the failure contribution probability of that path.
[0052] If a coupling failure relationship is detected between upstream systems during the actual simulation, the system defaults to a conservative approach, meaning that instead of multiplying probabilities together, the maximum value of the failure contribution probabilities of each upstream path is taken as the overall failure probability. Each agent adds the overall failure probability to its own state calculation in the next time step. The expansion of the power outage area drives the deterioration of each agent's state, and the changes in each agent's state, in turn, affect the power outage recovery conditions through the coupling matrix. This iterative process continues until the preset total simulation time is reached or all systems reach a stable final state.
[0053] 7. Identification of critical time windows.
[0054] This step is calculated synchronously during the execution of steps five and six.
[0055] The calculation method for the golden window deadline T3 is as follows: During the cascading failure simulation, the system synchronously calculates a judgment function that determines whether "if all available resources are deployed at this moment, the final power outage ratio can be reduced to below a preset threshold". The system maintains a dynamic ledger of schedulable resources during the simulation process. This ledger is deducted in real time with resource allocation events and restored in real time with replenishment events. The "total remaining schedulable resources" in the judgment function is directly read from this ledger.
[0056] The decision function estimates the power outage area based on the current accumulated outage area, the current diffusion rate, and the total remaining schedulable resources. The specific calculation formula is: the predicted final outage ratio equals the current outage ratio plus the diffusion rate multiplied by the characteristic response time divided by the total city area, minus the resource reduction ratio. The characteristic response time is the average delay from the issuance of the resource call command to the actual effect; this parameter defaults to the average statistical value of the city's historical power outage emergency responses. If no historical data is available, the empirical default value of fifteen minutes is used, which users can adjust in the scenario parameters. The resource reduction ratio is estimated based on the total remaining schedulable resources and the historical average reduction effect per unit of resource. When the predicted final outage ratio changes from less than a preset threshold to greater than a preset threshold, the corresponding time point is the golden window deadline T3. This estimation is completed synchronously in a single simulation, eliminating the need to repeat the complete simulation for each potential intervention moment.
[0057] The power outage trigger time T0 is directly obtained from the initial fault time input by the user. The first cascading fault initiation time T1 is obtained by scanning the protection action event sequence, recording the time point when the newly added equipment trips due to the first overload protection action after the initial fault. The first secondary disaster initiation time T2 is obtained by scanning the state change records of each infrastructure agent, recording the earliest of the following three conditions: the traffic congestion index first exceeds twice the normal value, the communication base station first goes offline due to power outage, or the water supply pressure first falls below the minimum service head.
[0058] The critical infrastructure batch failure time T4 is calculated by weighted summing of the failure states of the five actively propagating systems in the coupling correlation matrix. The weights here are static prior weights reflecting the importance of each system in the power outage disaster, pre-calibrated by domain experts based on urban operational mechanisms. A binary failure state of the medical system is overlaid as an independent additional term; that is, when the weighted sum first exceeds 0.7 or the medical system fails, the earlier of these two times is determined as T4. The recovery start-up time T5 is backfilled after step ten. After all critical nodes are calculated, the system generates an annotated timeline, displaying the name and specific time at each node location.
[0059] 8. Disaster impact assessment.
[0060] The power outage area polygon is spatially overlaid with a population distribution grid. For each grid, the proportion of its area falling within the power outage area is calculated, multiplied by the population of that grid, and then summed to obtain the total affected population. Simultaneously, age structure data is weighted to calculate the number of vulnerable groups such as the elderly and children. Critical facility assessment involves traversing the geographical coordinates of facilities such as hospitals, schools, and transportation hubs to determine if they are located within the power outage area. If so, the backup power configuration and remaining operating time of the facility are further queried, outputting the power outage status level and estimated failure time for each facility.
[0061] 9. Emergency resource allocation simulation.
[0062] This step runs in parallel throughout the entire deduction process from steps five to eight.
[0063] First, a resource database is constructed, importing the real-time location and availability status of emergency repair teams, emergency power generation vehicles, emergency communication vehicles, flood control materials, and medical emergency forces from various business systems. Each type of resource is stored as a record with spatial coordinates and attribute fields, and each resource type additionally records its duration of validity.
[0064] Upon receiving a resource allocation request, the system extracts parameters such as the required location, type and quantity of resources, and required arrival time. It then filters available resources from the resource database that meet these criteria. For each candidate resource, the system invokes a route calculation service. This service uses real-time traffic conditions provided by a traffic digital twin model as input and a dynamic shortest path algorithm to calculate the recommended route and estimated travel time from the resource's current location to the required location, aiming to minimize travel time. After calculation, all candidate resources are sorted by estimated arrival time, and the optimal resource combinations are selected to form the allocation plan.
[0065] Resource bottleneck area identification involves dividing the city into several analysis grids. For each grid, the ratio of the power outage severity score to the surrounding resource coverage score is calculated. The grid with the highest ratio is marked as a resource bottleneck area, and a list of available resources in the surrounding area and suggested allocation routes are output. The power outage severity score is a weighted sum of three indicators: the percentage of power outage area within the grid, the affected population density, and the number of critical facilities. The weights are set by the user in the system configuration.
[0066] Once emergency resources arrive at their designated locations and are put into use according to the allocation plan, their recovery effect is fed back in real time to the status calculation of each infrastructure agent in step six. After the resources are deployed, the corresponding functional indicators of the target area immediately jump to a certain percentage of the normal level. Thereafter, this recovery level is maintained constant for the continuous effective duration of the resources. When the continuous operation time of the resources exceeds the effective duration and no replenishment occurs during this period, the recovery effect gradually degrades according to the preset decay rate until it reaches zero. If the system detects a resource replenishment event, the timing and amount of the replenishment event are preset by the user during scenario configuration as scenario policy parameters. After the replenishment event occurs, the remaining value of the effective duration is reset to the initial value and extended accordingly. In multi-scenario parallel simulation, each scenario holds an independent copy of the resource status. The initial resource status is copied from the real-time database, and subsequent consumption and replenishment evolve independently within each copy without affecting each other.
[0067] 10. Parallel simulation of multiple scenarios.
[0068] The system constructs a distributed cluster consisting of a master node and multiple computing nodes. The master node receives multiple simulation tasks submitted by users. Each task contains complete power outage scenario parameters and emergency strategy schemes to be evaluated. The master node assigns these tasks to idle computing nodes and monitors their execution progress. Each computing node runs an independent replica of the spatiotemporal evolution simulation engine. Each replica independently executes all calculations from steps five to nine. The replicas do not share memory and only synchronize their states through message passing.
[0069] The parallel simulation service has an embedded resource scheduling submodule, which is responsible for managing the resource allocation between formal simulation tasks and background optimization tasks in the computing node pool. Formal simulation tasks have a higher priority than background optimization tasks. When a formal simulation task occupies all computing nodes, the optimization task enters the waiting queue. When node resources are released, the optimization task automatically resumes execution.
[0070] After all computing nodes complete their respective tasks, they return the results to the master node. The master node collects all results and performs a comprehensive comparison. First, it normalizes the results of each scheme on each evaluation index to eliminate dimensional differences. Then, it calculates the relative closeness of each scheme using the superior-inferior solution distance method. The weights of each evaluation index are objectively determined using the entropy weight method, which automatically calculates the weights based on the data dispersion of each scheme on each index. Indicators with greater dispersion are assigned higher weights, thus avoiding bias caused by subjective weighting. Finally, the scheme with the highest closeness is output as the optimal recommendation strategy.
[0071] The entropy weights used here serve different purposes from the expert prior weights calculated in step seven (T4). The former characterizes the significant differences in evaluation indicators among the various simulation schemes and is a dynamic, data-driven weight; the latter characterizes the inherent importance ranking of each infrastructure system in a power outage disaster and is a static, prior weight. Mathematical consistency between the two is not required. After the strategy simulation is completed and the optimal scheme is selected, the system automatically fills the timeline annotation generated in step seven with the time when the power grid begins to restore power, as determined in the recovery strategy corresponding to that scheme, as the recovery start time T5.
[0072] 11. Visual output.
[0073] This step is executed uniformly after all the multi-scenario parallel simulations in Step 10 are completed. The report generation service extracts all key data from the database for this simulation, fills in the content according to the preset report template, and sequentially generates a disaster evolution summary section, a chart annotating key time windows, a time-segmented impact assessment statistical table, a multi-scheme comparison matrix, and a description of recommended strategies and solutions, outputting it as a document format file.
[0074] The 3D visualization service is presented on the user interface. After the system loads the city 3D model scene, it receives the time-series data output by the inference engine in real time. It generates a semi-transparent red overlay layer based on the power outage boundary data at each moment and dynamically updates its shape. It generates corresponding color marks and statistical chart panels based on the status data of each infrastructure intelligent agent. Users can rotate the view and zoom the view by dragging the mouse in the 3D scene, and click on any facility object to pop up its detailed status information window.
[0075] This invention also provides a dynamic simulation system for implementing the above methods. This system comprises ten microservice modules that interoperate via application programming interfaces (APIs) and remote procedure call protocols. The simulation engine employs a variable time step strategy: microsecond-level steps are used for the electromagnetic transient simulation phase, millisecond-level steps for the electromechanical transient phase, second-level steps for the cascading fault power flow iteration phase, and a fixed one-minute step for the secondary disaster evolution and crowd behavior simulation phases. The electromagnetic transient simulation terminates after completing the protection action judgment, and the system automatically switches to the electromechanical transient simulation. The electromechanical transient simulation automatically switches to the cascading fault power flow iteration phase after the system frequency and voltage fluctuations decay to within the allowable steady-state range. The secondary disaster evolution step is fixed at the minute level and does not depend on the end of the power grid transient process. During data access, various data types are automatically aligned to the time corresponding to the current step size by selecting the nearest sampled value or using linear interpolation based on the current simulation phase.
[0076] The key dependencies between the microservices are as follows: Before starting the formal simulation, the spatiotemporal evolution simulation engine must confirm that the twin modeling service has completed loading of the complete state profile and returned a ready signal; the visualization service independently subscribes to the simulation result topics in the data bus, and is asynchronously decoupled from the main computation process; the data access service is responsible for receiving various raw data streams, performing quality control and format conversion, and then writing them to the data bus; the twin modeling service is responsible for maintaining the topology and state variables of the digital twin model; the scenario management service is responsible for storing and retrieving power outage scenario parameter templates and historical case data; the spatiotemporal evolution simulation engine, as the computational... The computationally intensive module implements the core algorithm using a compiled language and deploys it on GPU-accelerated computing nodes; the crowd behavior simulation service runs a multi-agent simulation framework independently and exchanges data with the main engine through a message queue; the resource simulation service maintains the resource database and performs path optimization calculations; the parallel inference service is responsible for task distribution and result aggregation and sorting, and its embedded resource scheduling submodule is also responsible for managing the resource allocation between formal inference tasks and background optimization tasks in the computing node pool; the visualization service undertakes 3D scene rendering and human-computer interaction functions; and the report generation service automatically extracts data and fills in templates after the inference is completed.
[0077] All microservices are deployed using a containerized approach, supporting dynamic adjustment of the number of instances based on load to adapt to the application needs of cities of different sizes. Regarding service scaling strategies, the stateful inference engine service does not participate in automatic scaling during task execution. After the task is completed, the engine first writes the inference results to local temporary storage, then reports the task completion to the scheduler and enters a releaseable state. At the same time, an independent background data archiving service asynchronously migrates the results from local temporary storage to persistent storage. The engine container can be released without waiting for archiving to complete. Stateless services can scale arbitrarily.
[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A method for simulating and extrapolating large-scale power outage scenarios in cities based on digital twins, characterized in that: include: Acquire multi-source operational data of the city; Construct a digital twin model of infrastructure, wherein the digital twin model of infrastructure includes a power grid sub-model, a transportation sub-model and a water supply sub-model, wherein the power grid sub-model, the transportation sub-model and the water supply sub-model are coupled and related. Obtain the triggering parameters for power outage scenarios; based on the multi-source operational data and the digital twin model of the infrastructure, simulate and extrapolate the triggering parameters for power outage scenarios to obtain the disaster simulation and extrapolation results and key time windows; The results of disaster simulation and key time windows are visualized.
2. The method according to claim 1, characterized in that, The multi-source operational data includes power grid operational data, spatial data, meteorological data, traffic operational data, base station operational data, and water and gas supply operational data.
3. The method according to claim 1, characterized in that, In the digital twin model of the infrastructure, the power grid sub-model uses generators, buses, and loads as node objects and lines as edge objects, and generates a topology structure based on the node objects and edge objects, and sets a state estimator based on the topology results; the traffic sub-model is constructed by dividing the corresponding cells according to road length and traffic capacity; and the water supply sub-model is constructed based on the network composed of water supply pipeline nodes and pipe segments.
4. The method according to claim 1, characterized in that, The coupling correlation between the power grid sub-model, the transportation sub-model, and the water supply sub-model is established based on the coupling correlation matrix; wherein the coupling correlation matrix includes coupling elements between different systems; wherein the different systems include power grid, transportation, communication, water supply, and gas systems; and the coupling elements include strength coefficient, propagation probability, and time delay.
5. The method according to claim 1, characterized in that, The process of simulating and extrapolating the disaster triggering parameters for power outage scenarios includes: Obtain the triggering parameters for power outage scenarios; based on the multi-source operational data and the infrastructure digital twin model. The multi-source operational data is used as the initial state of the infrastructure digital twin model; the parameters of the infrastructure digital twin model in the initial state are adjusted by triggering parameters; transient and steady-state fault simulations are performed using the adjusted infrastructure digital twin model; during the fault simulation process, secondary disaster chain evolution simulation, critical time window identification, and emergency resource allocation simulation are performed to obtain disaster simulation and deduction results and critical time windows.
6. The method according to claim 5, characterized in that, The process of simulating the evolution of the secondary disaster chain includes: Within each simulation time step, the traffic agent, communication agent, and water supply agent in the infrastructure digital twin model adjust their states. At the end of each time step, each agent writes its own state into the coupling correlation matrix and reads the states of other systems, and obtains the failure probability. The failure probability is calculated based on the failure contribution probability of the corresponding upstream path. Each agent adds the comprehensive failure probability to the state calculation of the next time step.
7. The method according to claim 5, characterized in that, Key time windows include: The timeframes for power outage triggering, initial cascading failure initiation, initial secondary disaster initiation, golden window cutoff, critical infrastructure mass failure, and recovery restart are determined. The initial secondary disaster initiation time is obtained by predicting the final power outage ratio. The golden window cutoff time is obtained by weighted summation of the failure states of the five active propagation systems in the coupling correlation matrix and threshold determination. The critical infrastructure mass failure time and recovery restart time are obtained by backfilling the time when the power grid begins to restore power supply in the optimal solution.
8. The method according to claim 5, characterized in that, The process of emergency resource allocation simulation includes: A resource database is constructed. When a resource allocation request is triggered, available resources are filtered from the resource database based on the triggered resource allocation request. Using real-time traffic conditions provided by the traffic digital twin model as input, a dynamic shortest path algorithm is used to calculate the recommended path and estimated travel time. The optimal resource combination is selected according to the estimated arrival time to form an allocation plan. The allocation plan is simulated to obtain the recovery effect, and the recovery effect is fed back to the state calculation of each infrastructure agent in real time.
9. A simulation and deduction system for large-scale urban power outages based on digital twins, characterized in that: Used to perform the method described in any one of claims 1-8.
