A Power Grid Risk Early Warning Method Based on Digital Twins

Abstract

This invention discloses a power grid risk early warning method based on digital twins, belonging to the field of power grid risk early warning technology. The method includes: firstly, integrating data from the power grid energy management system, production management system, and information security system to construct a digital twin base for the primary system physical model, the secondary system mechanism model, and associated operating modes; then, considering the virtual automatic voltage control constraint calculation baseline situation; subsequently, injecting hypothetical equipment faults, and linking the execution of electromechanical transient analysis, relay protection action prediction, standby automatic transfer action prediction, and power flow recalculation to form a predictive topology; further, combining power source backtracking and dynamic voltage difference calculation at both ends after virtual disconnection of tie switches, identifying single-source risks, risks from the same upstream power source, and risks of double circuits on the same pole. This method achieves unified extrapolation of power grid operating status, secondary device actions, and structural risks, improving the completeness, foresight, and targeted level of risk identification and response.

Description

Technical Field

[0001] This invention relates to the field of power grid risk early warning technology, specifically a power grid risk early warning method based on digital twins. Background Technology

[0002] The main regional power grid and urban power grid dispatching scenarios are applicable to networks with a wide variety of equipment, frequent switching of operating modes, a large number of users, and numerous interconnection and transfer operations, as well as loop closing and disconnection operations. In situations such as maintenance outages, load transfers, bus voltage loss, and backup power switching, dispatchers need to continuously monitor equipment status, topology changes, and the impact of power outages. Currently, they mainly use power grid models, real-time operating information, topology, accidents, power flow, risks, and classifications to assist in the arrangement of operating modes and risk prediction.

[0003] Chinese patent document CN108596450B discloses a power grid risk early warning method and system. First, it obtains the operating and structural parameters of transformers, circuit breakers, protection devices, and automatic transfer switches (ATS). Then, it inputs these parameters into a risk analysis system composed of a data processing layer, a risk analysis layer, and an early warning display layer. The risk analysis layer can establish a topology analysis model. When a fault occurs, it determines whether a circuit breaker will operate based on the protection device configuration. When a fault is isolated, it determines whether the circuit breaker triggered by the ATS will operate based on the protection action and ATS configuration. Then, it performs topology analysis on the power grid, identifies the electrical islands caused by the fault, and modifies the topology analysis model. The system also discloses a load analysis unit that calculates the load reduction amount of different nodes according to the load shedding sequence and range, and constructs a load reduction model. A node cluster growth algorithm is disclosed to analyze the substation topology and system topology and obtain risk values ​​under risk scenarios. Finally, the display layer publishes early warning information.

[0004] Under the aforementioned technical approach, existing technologies can directly conduct risk analysis on fault scenarios, protection actions, automatic transfer switches, and load shedding. However, they are not successful for the complex operating scenarios of regional and urban power grids. Firstly, the models, measurements, equipment ledgers, and protection information sources of various business systems are fragmented, have different refresh cycles, and inconsistent object identification standards. While existing methods can extract equipment characteristic parameters, they rarely explain how to stably characterize the relationship between primary topology state, secondary action state, and subsequent load consequences at the same point in time. Secondly, in operating modes where main transformer voltage regulation, reactive power compensation activation / deactivation, interconnection power transfer, and backup power switching coexist, the electrical states of the pre-fault section, voltage and reactive power distribution, post-fault protection and automatic transfer switch chains, and the electrical states before and after interconnection operations are coupled. Segmenting these related links leads to risk analysis focusing primarily on interruption consequences, load shedding, or comprehensive risk values, making it difficult to provide consistent judgments for different operating scenarios.

[0005] Therefore, how to continuously describe the consequences of faults, topology, and power supply under a unified operating scenario under complex operating conditions is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a power grid risk early warning method based on digital twins. It constructs a digital twin base for a primary system physical model, a secondary system mechanism model, and associated operating modes; then it considers the virtual automatic voltage control constraint calculation baseline situation; subsequently, it injects hypothetical equipment faults, and performs coordinated electromechanical transient analysis, relay protection action prediction, standby automatic transfer switch action prediction, and power flow recalculation to form a predictive topology; further, it combines power source tracing and dynamic voltage difference calculation after the virtual disconnection of tie switches to identify single-source risks, risks from the same upstream power source, and risks of double circuits on the same pole. This method achieves a unified deduction of power grid operating status, secondary device actions, and structural risks; and solves the technical problems described in the background art.

[0007] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: The power grid risk early warning method based on digital twins includes: acquiring the CIME model and real-time cross-sectional data of the power grid energy management system (EMS), the equipment ledger of the production management system (PMS), and the real-time secondary protection data of the information protection system; constructing a digital twin base for the primary system physical model, the secondary system mechanism model, and associated operation modes; calculating the baseline situation based on the digital twin base and considering virtual AVC constraints; injecting hypothetical equipment faults into the digital twin base, performing transient analysis, relay protection action prediction, backup automatic transfer action prediction, and power flow recalculation to obtain the predicted topology; Based on the predictive topology, single power source risk, risk of the same upstream power source, and risk of double circuit on the same pole are identified, and the dynamic voltage difference between the two ends is calculated after the tie switch is virtually disconnected. The absolute load reduction and load transfer are quantified according to the load distribution before and after the fault, and an event level and power grid risk warning notice are generated accordingly.

[0008] Furthermore, when constructing the digital twin base for the accompanying operation mode, object association, parameter identification, and data cleaning are performed on the CIME model, equipment ledger, and real-time secondary protection data. After establishing the correspondence between primary equipment and secondary devices, a unified object identifier is generated.

[0009] Furthermore, the accompanying operation mode digital twin base also includes an operation information model constructed based on real-time cross-sectional data and historical trends. The operation information model uniformly writes the current topology status, measurement status, and equipment commissioning / decommissioning status, and serves as the input cross-section for subsequent baseline situation calculations.

[0010] Furthermore, the baseline situation for virtual AVC constraint calculation is taken into account, including constructing an admittance matrix based on the bus, line, main transformer, parallel reactive power equipment and switch status in the primary system physical model, and simultaneously taking into account the main transformer tap changer action sequence and reactive power compensation resource commissioning and decommissioning constraints during the power flow iteration process.

[0011] Furthermore, after injecting hypothetical equipment faults into the digital twin base, a fast-process electromechanical transient analysis is first performed, and the first round of relay protection actions and corresponding isolation ranges are predicted based on the protection setting area, pressure plate status, output circuit and circuit breaker status in the secondary system mechanism model.

[0012] Furthermore, after the first round of relay protection actions causes topology changes, the process of predicting topology formation further includes: calling virtual AVC constraints to update the voltage distribution of the entire network, predicting automatic transfer switching actions based on the power failure side status, backup power supply side status, and backup switch status, and performing power flow recalculation based on the action results.

[0013] Furthermore, when identifying single-power-source risks, risks from the same upstream power source, and risks of dual circuits on the same pole, the power supply sources of the busbars and feeders are traced back along the energized connection relationships in the predicted topology, and the corresponding risk types are determined by combining the line corridor information and pole channel information in the equipment ledger.

[0014] Furthermore, when calculating the dynamic voltage difference across the tie switch, while keeping the rest of the network environment in the predicted topology unchanged, the tie switch is virtually disconnected in the digital space, the voltage state of the busbars across the tie switch is recalculated, and the dynamic voltage difference is compared with a preset threshold.

[0015] Furthermore, when quantifying the absolute load reduction and load transfer, the load distribution status before the hypothetical equipment failure and the load distribution status after the completion of the full sequence operation are recorded. The continuous power loss load and the transferred load are distinguished by the user access point as the statistical object, and the power loss status of important users is extracted.

[0016] Furthermore, when generating event levels and power grid risk warning notices, the absolute load reduction value, power outage status of important users, automatic transfer actions of backup power supplies, and details of load reduction are sent to the evaluation rule tree engine. Based on the load reduction threshold and the bus full shutdown condition, level 5, level 6, or level 7 events are determined, and risky substations and lines are marked through the RPC framework linked to the front-end dynamic three-dimensional geographic interactive interface.

[0017] (III) Beneficial Effects This invention provides a power grid risk early warning method based on digital twins, which has the following beneficial effects: By centralizing the data of the power grid energy management system, production management system, and protection and information system on the digital twin base of the associated operation mode, a physical model of the primary system and a mechanism model of the secondary system are established, so that the equipment status, topology relationship and protection information are consistent on the same object chain, providing a unified data source for subsequent baseline situation calculation and fault inference.

[0018] By incorporating synchronism and virtual automatic voltage control constraints into the baseline situation calculation, and integrating the main transformer tap changer and reactive power compensation resources into the power flow iteration, the node voltage and reactive power distributions are made more consistent with the actual voltage regulation process, providing an initial situation for hypothetical equipment fault simulation. By adding hypothetical equipment faults to the digital twin base, fast-process electromechanical transient analysis, relay protection action prediction, standby automatic transfer action prediction, and power flow recalculation are performed sequentially, continuously completing the prediction of electrical disturbances, secondary device actions, and topology changes.

[0019] By tracing back the power supply source based on the predicted topology, and simultaneously calculating the dynamic voltage difference between the two ends after the tie switch is virtually disconnected, the risks of single power supply, same upstream power supply, double circuit on the same pole, and loop over-limit circulating current are identified in a unified manner, thus unifying the judgment of structural weaknesses and operational risks.

[0020] By comparing the load status before and after the fault, as well as the absolute load reduction and load transfer, and by incorporating the results of power outages of important users and automatic transfer of backup power, the event level determination is based on the final power supply consequences, making Level 5, Level 6, and Level 7 events more suitable for on-site handling.

[0021] The system generates power grid risk warning notifications and links them with the front-end dynamic 3D geographic interface through a remote process call framework. It marks substations, lines and related risk objects, converts the simulation results into visual handling information, and minimizes the transmission time from risk identification to dispatch and handling. Attached Figure Description

[0022] Figure 1 This is a diagram illustrating the overall architecture of the power grid risk early warning system based on digital twins according to the present invention. Figure 2 This is a schematic diagram illustrating the construction of the twin power grid base and cross-domain relationships of the present invention; Figure 3 This is a schematic diagram illustrating the solution of the normal operating baseline state considering virtual automatic voltage control constraints in this invention; Figure 4 This is a schematic diagram of the primary and secondary cascade deduction process under the hypothetical fault scenario of the present invention; Figure 5 This is a schematic diagram illustrating the power source tracing and differential voltage determination of the interconnection switch in this invention. Figure 6This is a schematic diagram illustrating the process of generating load reduction statistics, event classification, and early warning notifications according to the present invention. Detailed Implementation

[0023] 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.

[0024] Please see Figures 1-6 This invention provides a power grid risk early warning method based on digital twins, including: The CIME model and real-time cross-section from EMS, the equipment ledger from PMS, and the real-time secondary protection data from the information protection system are converged into a single set of computable, traceable, and continuously transferable twin power grid base. This ensures that subsequent power flow solutions, virtual AVC constraint injection, and protection and backup automatic transfer action simulations are all based on the same object, the same time axis, and the same parameter caliber.

[0025] First, a primary system physical model is established, followed by a secondary system mechanistic model. Then, historical power flow and current cross-section data are integrated into the operational information model, ensuring that three key facts—what the equipment is, how it is currently connected, and under what conditions it operates—are synchronously defined in the digital space. This co-operational mode not only preserves the topology and parameter boundaries of the physical power grid but also retains the triggering relationships between secondary and primary equipment. This provides a stable input for constructing the admittance matrix in step two and a directly accessible object index for fault injection and cascading action simulation in step three.

[0026] Step 1: Generate a twin power grid base with the same source, topology, and action semantics as the physical power grid. Regional power grids face challenges in dispatching and operation, including dispersed model sources, inconsistent naming conventions, and varying equipment status refresh cycles. If EMS cross-sections are used directly to drive power flow solutions, while primary equipment parameters can be included in the calculation, protection circuit breakers, setting zones, monitoring and control interlocking relationships, and automatic transfer switch (ATS) activation status cannot be received on the same time axis. Furthermore, if information protection is read separately, accurate mapping between the protection information and the busbars, lines, main transformers, and bays will be lost.

[0027] EMS provides busbar, line, main transformer, and switch status and cross-sectional measurements; PMS provides equipment ledger, bay assignment, line corridor identification, tower passage identification, and equipment limits; the protection information system provides protection setting zone number, pressure plate status, action position, alarm position, and CT / PT information; the historical power flow database provides historical segment vectors. The rule base provides voltage control zone targets, voltage threshold tables, permissible differential pressure threshold tables, and event level tables.

[0028] Therefore, we first determine whether the objects are the same object, then determine what computable state the object is in at the current moment, and finally write the state into the twin grid base.

[0029] The following actions are all performed collaboratively by the data processing server and the simulation computing cluster. The data processing server is responsible for data collection, cleaning, normalization, mapping, and missing data completion, while the simulation computing cluster is responsible for parameter identification, topology verification, and runtime information consolidation. Output results are uniformly written to the base library, which retains only four types of objects: primary equipment objects, secondary device objects, cross-domain associated objects, and cross-section status objects. These four types of objects carry equipment identifiers, connection relationships, action conditions, and timestamps, respectively, ensuring that no further field translation is required when directly reading data in step two.

[0030] First, the data processing server reads information from the EMS (Electronic Management System) regarding substations, busbars, lines, main transformers, switches, disconnectors, and measurement sections; from the PMS (Property Management System) it reads equipment ledgers, commissioning status, maintenance status, nameplate parameters, and associated bays; and from the protection information system it reads protection device numbers, setting zone numbers, pressure plate status, operating positions, and alarm positions. Then, using the substation code, bay name, primary wiring relationship, and voltage level as the primary index, and the equipment alias dictionary and historical mapping table as secondary indexes, it performs same-name merging and different-name merging on the three data sources. Objects that still cannot be uniquely located after merging are not directly deleted but enter a verification queue for further review based on topology relationships and electrical ratings. The mapping and matching process can be represented as follows: In the formula, the mapping matching degree : No. The first EMS / PMS object and the first The pairing strength of each guarantee object, with a value range of [value range missing]. Used to determine whether to establish cross-domain association objects; topological relationship consistency. The range of values ​​is This is used to characterize the adjacency consistency of two objects in the plant-bay-switcher chain; it is obtained by weighting plant consistency, bay consistency, adjacent bus consistency, and adjacent switch consistency; name semantic consistency. The range of values ​​is This is used to characterize the text proximity of device names, interval names, and aliases after dictionary normalization; it is obtained from the edit distance or word overlap rate of the standardized device names. Electrical rating conformity The range of values ​​is It is used to characterize the consistency of voltage level, capacity, CT / PT ratio, and wiring side information; it is calculated from the consistency of rated voltage, equipment type, CT ratio, and PT ratio; weighting coefficients. Weighting coefficients Weighting coefficients All values ​​are positive and satisfy the following condition: It is used to constrain the contribution ratio of the three types of evidence in the pairing determination. Initially, they are equally weighted. If there are labeled historical samples, offline correction is performed with the goal of minimizing the mapping error of the samples.

[0031] In one implementation, after maintenance personnel switch the maintenance status of a 110kV line in the morning, the PMS first updates the maintenance identifier in the equipment log, and the EMS then updates the switch position. Simultaneously, the protection information system displays the corresponding pressure plate exit information. The data processing server does not treat these three changes as three independent events, but rather merges them into a single cross-domain associated object, causing the line to simultaneously exhibit primary-side exit, secondary-side blocking, and power flow channel disconnection within the twin space. This prevents subsequent steps from exhibiting the bias of the protection system still being treated as operational devices even though the line has undergone maintenance.

[0032] Preferably, parameter identification performs constraint back calculations on the main transformer tap changer, line impedance, and bus section status; if a complete CIME incremental file cannot be directly provided on site, a baseline topology is first generated based on the most recent full model, and then incremental objects are supplemented using the cross-section switch positions and ledger change records.

[0033] Therefore, the physical model of the primary system and the mechanism model of the secondary system share the same object primary key, and subsequent protection actions, automatic switching condition judgment and power flow calculation can be carried out around the same device, avoiding repeated renaming of the same object in the previous and subsequent steps.

[0034] After object normalization is completed, the data processing server sends the power, voltage, current, switch position, tap position, and historical power flow segments of the current section into the operational information model. The operational information model does not directly overwrite the original section; instead, it first determines whether there are sampling gaps, timescale drift, or local measurement abrupt changes in the current section, and then decides the fusion ratio between the current section and historical segments. Its purpose is to ensure that even when the physical power grid experiences short-term communication interruptions, single-station telemetry freezes, or delayed arrival of information transmission messages, it can still write continuously calculable associated states to the twin power grid base. The fusion process can be represented as follows: In the formula, the associated state vector :time The final operating state of the twin grid base is written, with the value range limited by the physical boundary of the measurement, and is used to output a unified cross-section to step two; real-time cross-section vector. :time The status set directly read from EMS and the information security system is used to carry the real-time status of the current substations and lines; historical fragment vectors Historical trends and action segments similar to the current operating mode are used to supplement continuity when there are local missing measurements. Fusion coefficient The range of values ​​is This is used to control the writing ratio between the current section and historical segments; base reliability. The range of values ​​is The result is obtained by summarizing the aforementioned object mapping results and is used to characterize the completeness of the current base at the object level; missing test penalty. The range of values ​​is Used to characterize time The degree of cross-sectional missingness, timescale drift, and abnormal jumps.

[0035] Base credibility Take the average matching degree of all valid cross-domain associated objects at the current moment; missing test penalty. It consists of three parts: measurement missing rate, timestamp drift rate, and abnormal jump rate measurement rate; real-time cross-section vector. Messages from EMS and the Information Security Management System at the current moment; historical fragment vectors From a historical cross-section database categorized by operating mode.

[0036] In one implementation, during the evening peak hours, telemetry messages from a substation experience a brief pause, while those from adjacent substations and transmission lines continue to arrive. The operational information model first retains the most recent reliable state of the substation, and then writes historical power flow segments under the same operating mode into the associated state vector according to the above formula. The data processing server then marks the station object as a section to be filled in. When a new message is recovered, the data processing server only replaces the local state of the station object, without rewriting the remaining objects. This results in a continuous operating mode that remains visually consistent and allows for the construction of a complete admittance matrix in computation.

[0037] As a parallel expansion scheme, historical fragment vectors It can be output from a neural network, such as a gated recurrent unit network, or retrieved from a sample library established according to season, load type, and switching mode; neither path changes the main chain of first object normalization, then state assembly, and finally writing to the base library. Thus, the operational information model is no longer an isolated prediction unit, but rather constitutes a writing mechanism for the associated operational mode together with the primary system physical model and the secondary system mechanism model.

[0038] In use, through the two interconnected processing steps described above, the twin grid base outputs not only the visible topology but also a unified combination of equipment primary keys, operating conditions, cross-sectional states, and time indices. Therefore, step two can directly extract the admittance matrix and virtual AVC constraint objects based on this. Simultaneously, protection devices, monitoring and control devices, and automatic transfer switches are locked to specific primary equipment and bay links. In step three, when a hypothetical fault is injected, the corresponding mechanism judgment can be directly triggered without any drift in the action entity.

[0039] Furthermore, the accompanying state vector This ensures that cross-sections with localized missing measurements remain continuous and calculable, thus providing a stable starting point for subsequent risk quantification, ring network differential pressure calculation, and load reduction statistics.

[0040] Step 2: Establish a normal operating baseline state considering virtual AVC constraints on the twin power grid base. Embed the virtual AVC system into the power flow iteration loop, and instead of leaving AVC actions to be repaired after power flow solution, update the control area target, voltage deviation, and reactive power output allocation synchronously during the solution process.

[0041] Based on the primary system physical model output in step one, extract the states of the power plant, busbars, lines, main transformers, parallel capacitor banks, parallel reactor banks, and switches, and construct the complex admittance matrix. Simultaneously, real-time load and power generation output of the entire network are extracted to generate active power injection vectors and reactive power injection vectors at nodes; then, virtual AVC constraints are introduced within the Newton-Raphson iterative framework to obtain the normal operating baseline state.

[0042] The associated state vector output in step one The substations, lines, main transformers, protection devices, and automatic transfer switches have been incorporated into a unified system. However, this state is still merely a composite result of the current wiring and measurements. If it is directly fed into the conventional power flow program, the node voltages and reactive power distribution can only reflect the static cross-section and cannot reflect the linkage and constraints between the main transformer tap changers, parallel capacitor banks, parallel reactor banks, and generator excitation reactive power regulation within the voltage regulation control area. This is precisely where deviations are prone to occur in the power grid: while the voltages of each bus may appear within the allowable range on the screen, the actual control chain is already approaching the reactive power regulation limit. If subsequent fault disturbances are superimposed, the risk will be significantly amplified.

[0043] The following actions are performed collaboratively by the data processing server, simulation computing cluster, and control policy library. The data processing server starts from the associated state vector. The system extracts node injection power, switch position, tap position, and reactive power equipment activation / deactivation status, and generates a complex admittance matrix accordingly. The control strategy library reads the target voltage table, main transformer voltage regulation sequence table, reactive power compensation equipment activation / deactivation sequence table, and generator reactive power boundary table for each voltage control zone. The simulation calculation cluster merges the above two types of information into a single residual equation within the Newton-Raphson iterative framework.

[0044] Therefore, the output of step two is not only the node voltage after the power flow calculation, but also a unified baseline situation including control area affiliation, voltage regulation sequence, reactive power margin and node status, which can be directly inherited by step three.

[0045] The data processing server first rearranges the bus order according to the balancing node, generating node, and load node, and then assembles the complex admittance matrix based on the line parameters, main transformer short-circuit impedance, tap changer ratio, bus connection relationship, and switch conduction status in the primary system physical model. In this context, the disconnected switch is not left in the matrix as a zero-impedance object, but is directly removed from the branch adjacency relationship; the main transformer tap position is not replaced by a fixed correction value, but is written into the branch transformation ratio coefficient, so that this coefficient can change in conjunction with the virtual AVC constraint in subsequent iterations. Within the power system domain, the complex admittance matrix... The assembly is a mature and well-known technology. Those skilled in the art can implement it based on line impedance, main transformer ratio, parallel branch parameters, and switch states. Complex admittance matrix The parameters are obtained by assembling the series admittance of energized branches, parallel admittance, main transformer tap ratio, and parallel branch parameters according to the busbar node method.

[0046] Subsequently, the simulation computing cluster, based on the associated state vector Generate node state vectors Node state vector It consists of the voltage magnitude and phase angle of each non-equilibrium node, serving as the sole update object for power flow iteration. To ensure the reliability of the base... In this step, the simulation computing cluster writes the control associations of the virtual AVC system on the current section as a node-resource mapping matrix. Then it is combined with the regular Jacobian matrix to form an augmented Jacobian matrix. : Where: conventional Jacobian matrix The node state vector is determined by the active power imbalance and reactive power imbalance. A matrix composed of the first-order analytic derivatives is used to provide the local linearization result of the conventional power flow equations, and the values ​​of its elements vary with the current cross section; base reliability. The range of values ​​is This is used to adjust the effect of the object completeness from step one on the constraint injection strength; when the object mapping is sufficient, the virtual AVC constraint can enter the solution loop more deeply; Node-Resource Mapping Matrix This is used to characterize the topological interaction relationships between reactive resources, main transformer tap changers, and controlled nodes within the control area. Its elements take values ​​of zero or non-zero real numbers; zero indicates no association, and non-zero values ​​indicate the direction and strength of the interaction. Node-resource mapping matrix. The rows correspond to the controlled nodes, and the columns correspond to the voltage regulation resources. When the column object is a main transformer tap, the matrix element takes the sensitivity of that tap to the voltage of the controlled node. When the column object is a generator, the matrix element takes the sensitivity of the generator's reactive power increment to the voltage of the controlled node. When the column object is a parallel capacitor bank or a parallel reactor bank, the matrix element takes the sensitivity of its reactive power switching on / off to the voltage of the controlled node. The sensitivity can be obtained by the finite difference method or the decoupling sensitivity method under the primary power flow base point.

[0047] transpose marker Node-resource mapping matrix Row and column swapping is performed to re-project resource-side coupling information back to the node side; constraint weight matrix : is a diagonal matrix used to characterize the priority order of voltage targets for each controlled node and the intervention weight of various reactive resources; its diagonal elements take positive values, the larger the value, the more preferentially the node deviation is pushed back into the target band; it is a diagonal matrix, and its diagonal elements are jointly determined by the magnitude of the node voltage deviation and the node priority.

[0048] Augmented Jacobian Matrix Used to replace the regular Jacobian matrix This allows for the simultaneous solution of voltage control constraints and network equations in subsequent iterations. For each controlled node and each voltage regulating resource The node-resource sensitivity matrix is ​​calculated using the single-resource perturbation method: in, Indicates at time Resource variables When retrieving the current value of a node The voltage amplitude; Representing resources The amount of a single step; when resources When the main transformer tap is connected... Indicates a tap position; when resources When it is a parallel capacitor bank or a parallel reactor bank, Represents a set of throwing and cutting actions; when resources When setting the generator reactive power setpoint, This represents a reactive power adjustment step size.

[0049] Constraint weight matrix Using a diagonal format: in, Represents a node Target voltage; Represents a node Voltage dead zone; This indicates the node priority weight. In each power flow iteration, resource availability is checked sequentially in a fixed order: main transformer tap changer - parallel capacitor bank / parallel reactor bank - generator reactive power. When the node voltage deviation exceeds the dead zone, high-priority resources that have not yet reached their upper or lower limits are prioritized. When a resource reaches its action limit, it is removed from the adjustable resource set, and its current state is retained without further updates.

[0050] In one implementation, when the dispatcher checks the power flow of a certain area's feeders at midday, the front-end screen shows that the bus voltage of two 110kV substations is too high, but no switching operation is immediately issued on-site. The data processing server first writes the current position of the main transformer tap changer, the status of the parallel capacitor bank, and the generator reactive power boundary into the node-resource mapping matrix for that area. The simulation computing cluster was then equipped with an augmented Jacobian matrix. In this way, the bus voltage in the twin space is no longer determined solely by the load and power injection, but is also constrained by the availability of voltage regulation resources. What is ultimately retained on the screen is not simply the result of a higher voltage, but rather the reference voltage distribution at which the current resources will remain after their intervention.

[0051] In the augmented Jacobian matrix After establishment, the simulation computing cluster uses node state vectors The system performs Newton-Raphson iterations for the core unknowns, and synchronously updates the target voltage of the control area, the remaining boundaries of reactive power resources, and the movable range of taps in each iteration. If a generator has reached its upper or lower reactive power limit, the data processing server removes the generator from the voltage control resource set and rewrites it as a constant reactive power injection object. If a main transformer tap reaches its mechanical end position, the control strategy library freezes the tap position, retaining only other reactive power resources to continue participating.

[0052] This avoids obtaining a baseline situation that is mathematically feasible but unexecutable in the field. The iterative update relationship is written as: Where: State correction vector : The node state vector of the current iteration round The correction is used to gradually approximate the state where the node voltage magnitude and phase angle are both satisfied by the network equilibrium and control objectives; the augmented Jacobian matrix. The inverse matrix is ​​used to convert the residual vector Mapped to state correction vector In engineering implementation, it is preferable to use sparse triangular decomposition to solve linear equation systems rather than explicit inversion, in order to reduce the memory consumption of large-scale networks. residual vector The residual vector is composed of active power imbalance, reactive power imbalance, and voltage deviation terms introduced by virtual AVC constraints. Its elements can be positive, negative, or zero. The larger the absolute value, the more significant the deviation of the current node from the network balance and control objective. It is obtained by concatenating the active power imbalance, reactive power imbalance, and voltage deviation of the virtual AVC controlled nodes. Iteration cycle state vector. : No. The node state of each iteration, and the state vector of each iteration. The node state after one correction is used to generate the residual vector for the next round. ; In one implementation, a main transformer in a 220kV substation is responsible for regional voltage regulation. The control strategy library pre-programs the action sequence as follows: first, adjust the main transformer tap changer; then, connect the parallel capacitor bank; and finally, activate the generator reactive power. The simulation cluster iterates through several rounds, first changing the main transformer's turns ratio, then determining whether the node voltage still deviates from the target voltage range. If it still deviates, the parallel capacitor bank's connection state is rewritten into the associated state vector. The iteration continues in derived copies. The front-end screen then displays the changes in tap position and the synchronous change in reactive power support path, showing maintenance personnel a complete voltage regulation chain, rather than isolated node voltage figures. As a parallel expansion scheme, if the network is large and local coupling is weak, the simulation computing cluster can also use a domain decomposition solver on the periphery, while retaining the same augmented Jacobian matrix within the region. and residual vector Define them so that the terminology and data interface remain unchanged.

[0053] When used, the complex admittance matrix Node-resource mapping matrix With the credibility of the base Placed within the same solution framework, the voltage distribution obtained in step two no longer deviates from the actual boundaries of the control resources. Node state vector Each correction simultaneously responds to network balancing and virtual AVC constraints, thus the output normal operating baseline can be directly used as the starting point for the fault injection in step three, and remains consistent with the field voltage regulation chain. Furthermore, the main transformer tap changers, parallel capacitor banks, parallel reactor banks, and generator reactive power boundaries are checked round by round, reducing the possibility that subsequent risk simulations are based on unenforceable states.

[0054] Step 3: Under the normal operating baseline, transform the hypothetical equipment failure into a continuously traceable primary and secondary cascaded action chain and a post-fault prediction topology. Inject hypothetical faults, perform fast-process electromechanical transient analysis, protection action discrimination, virtual AVC recalculation, automatic transfer switch logic verification, and slow-process power flow recalculation into a single chain, enabling the primary system physical model and the secondary system mechanism model to relay each other within the same fault scenario.

[0055] Step 2 gives the node state vector Complex admittance matrix and associated state vector This already reflects the network balance state before the fault occurred, but risk warning goes beyond simply assessing current safety; it addresses how the power grid will evolve after the fault. If only static data is collected... Although section removal can reveal changes in power flow in some branches, it cannot explain why a certain line was tripped by protection, why power supply was not immediately restored after a section of bus lost voltage, or under what conditions the automatic transfer switch was activated or failed to operate.

[0056] The following actions are executed sequentially by the data processing server, simulation computing cluster, and control strategy library. The data processing server first replicates multiple scenario sandboxes from the normal operating baseline state output in step two, with each scenario sandbox containing only one set of fault scenarios. The system records the fault object identifier, fault type, fault location, start time, disconnection time, and circuit breaker failure to operate flag. The simulation computing cluster then calculates the fast-process electrical response within each scenario sandbox, filters out the first round of protection actions based on the secondary system mechanism model, and then hands the post-action topology to the control strategy library. This triggers virtual AVC rebalancing and automatic transfer switch condition verification sequentially. Finally, it performs slow-process power flow recalculation on the predicted topology, forming the post-fault state for subsequent steps.

[0057] During this phase, the data processing server does not directly delete the faulty device. Instead, it first writes the faulty object as a network disturbance and then allows the simulation computing cluster to advance along a very short time step.

[0058] When a fault first occurs, whether the protection device activates, whether the CT / PT sampling value exceeds the activation threshold, and which setting range the distance protection or current protection falls into all depend on the network state before the fault and the fault location. The simulation calculation cluster first converts the line, bus, main transformer, or switch object into an injection operator based on the fault location, and then compares it with the complex admittance matrix obtained in step two. Superimpose to generate the fault admittance matrix Then, the node state vectors are combined. With the accompanying state vector Find the transient response vector after the fault begins. : Where: time stamp Derivation timing; complex admittance matrix Step 2 outputs the pre-fault network admittance relationship; fault admittance matrix. Injection of fault scenario set The network admittance relationship obtained later; Fault injection operator Set up a set of failure scenarios and fault location coefficient A mapping operator that converts disturbances into branch or node admittance; inserts additional admittance into the corresponding branch segment during line faults, inserts ground fault admittance into the corresponding node during bus faults, and retains the relevant switch conduction state unchanged in the case of circuit breaker failure to operate; three types of fault injection methods: line faults: the faulty line is injected according to the fault location coefficient. The system is divided into two segments: from the beginning to the fault point and from the fault point to the end. Fault admittance is introduced at the fault point. For bus faults, ground fault admittance is introduced at the target bus node. For transformer faults, the coupling branches of the primary and secondary sides of the transformer are disconnected, and the equivalent fault branches are written according to the fault type. If the scenario includes circuit breaker failure to operate, the corresponding circuit breaker state is not modified at the time of fault clearing, but it is kept closed and the failure protection chain continues to be triggered.

[0059] When the fault object is the line At that time, according to the fault location coefficient Total line impedance Divided into two parts: Insert a fault node between the two segments. The fault type determines the node Ground access fault admittance .

[0060] When the faulty object is the busbar, the fault admittance is directly connected at the corresponding busbar node. When the faulty object is the main transformer, the original branch of the main transformer is disconnected, and an equivalent fault branch is inserted on the corresponding side according to the internal fault location. Among these, It represents a set of fault scenarios, including at least the faulty device identifier, fault type, start time, disconnection time, and circuit breaker failure to operate flag; Indicates the fault location coefficient of the line; This represents the fault admittance.

[0061] Fault Scenario Set The fault description object within the scenario sandbox includes the faulty device identifier, fault type, start time, disconnection time, and circuit breaker failure to operate flag; fault location coefficient. The range of values ​​is It is used to characterize the position ratio of the fault point from the beginning to the end of the line; when the fault object is a bus or main transformer, it is fixedly mapped to the corresponding endpoint. Transient response mapping The fast-process solution executed by the simulation computing cluster is implemented by alternating iterative steps between the network algebraic equations and the associated state equations of the generator, motor, voltage transformer, and current transformer; transient response mapping. The simulation is obtained through fast time-domain simulation; the simulation objects include at least the generator electromechanical state, network algebraic equations, and voltage and current phasors at the protection installation point; the protection measurements are converted by CT / PT proportional conversion and the power frequency phasors are extracted by one-cycle discrete Fourier algorithm; the transient deduction adopts implicit trapezoidal integral or equivalent numerical integrator.

[0062] Node state vector Step 2 outputs the set of pre-fault node voltage amplitudes and phase angles; associated state vectors. The set of operational cross-sectional states solidified by steps one and two; transient response vector. The set of electrical quantities after the fault begins is used to compare each item with the protection start-up conditions in the secondary system mechanism model. In one implementation, a single-phase grounding fault scenario is written into the middle of a 220kV line. The data processing server writes the fault location into the scenario sandbox. The simulation computing cluster first calculates the current surge and voltage drop observed at the protection installation points at both ends, and then sends these quantities to the setting area judgment chain of the corresponding line protection. On the front-end screen, the color of the faulty line changes abruptly first, and then the positions of the circuit breakers on both sides of the fault change to open, and the relevant intervals switch to isolation state.

[0063] Subsequently, the data processing server, based on the pressure plate status, setting zone number, CT / PT connection relationship, trip output circuit, and protection coordination sequence in the secondary system mechanism model, processes the transient response vector. Perform conditional judgment and generate the first round of action sequence. If the scenario sandbox contains a circuit breaker failure flag, then the first round of action sequence... Instead of directly modifying the topology, the failure switch is kept closed, and the next round of output discrimination of the failure protection or backup protection is triggered.

[0064] When the first round of action sequence Once formed, the data processing server first rewrites the on / off states and device activation / deactivation states in the scene sandbox based on this data, generating a predicted topology. The prediction topology here This is not simply the result of line reduction, but includes the wiring relationships of buses that remain energized after fault clearance, de-energized buses, and isolated primary equipment. The control strategy library predicts the topology. First, the virtual AVC constraints from step two are invoked to reallocate the main transformer tap changers and reactive power resources, and then the automatic transfer switch action flag is verified. Among them, the self-throwing action marker The operation is only established when the busbar on the power failure side is de-energized, the backup power supply side is energized, the backup line is in hot standby connection, the backup switch is ready to close, and the synchronization or de-energization detection conditions are met. If any of the conditions are not met, the power failure state is retained and the slow process is recalculated.

[0065] In one implementation, a section of a busbar at a 110kV substation loses power due to a fault in the upstream line, triggering the first round of action sequence. First, the faulty side line is disconnected. The control strategy library then checks if the tie line is energized, if the tie switch is in the hot standby position, and if the busbar de-energization detection circuit is established. If the condition chain is complete, the tie switch on the front-end screen changes from open to closed, and the power-loss-side feeder is restored to energized status. If any link in the condition chain is missing, the screen retains the power-loss color block. After completing the above rewrite, the simulation computing cluster predicts the topology. The slow-process steady-state power flow calculation is recalculated to obtain the active power distribution of the branches after the fault, and the amount of heavy overload accumulation in long-distance transmission channels due to load transfer is quantified. : Where: Overload accumulation amount After fault clearing, virtual AVC reallocation, and backup automatic transfer verification are completed, the cumulative limit exceedance level of key transmission channels across the entire network is assessed; Index A single line or branch of a single main transformer within a transmission channel set; a transmission channel set In predicting topology The set of lines and main transformer branches participating in the heavy overload statistics; channel weights Take a positive value to represent the first... The importance of a transmission channel in the power supply level and transfer capacity is determined by the voltage level, the power supply range, and the load-bearing capacity of important users. Active power of the channel after the fault In predicting topology The first result obtained by recalculating the power flow through the slow process Active power transmission value of each transmission channel; allowable power of the channel : No. The maximum power capacity allowed for a transmission channel under its current equipment condition; taken from the continuous thermal stability limits or operating limits table under the current equipment condition; channel allowable power. Taken from the equipment thermal stability limits or scheduling operation quota tables in the PMS ledger; channel weight The function is configured offline based on three factors: voltage level, whether it serves important users, and whether it is located on a main power transfer channel, and is read directly during runtime. Positive value extraction function, used to accumulate only the portion exceeding the limit; Thus, step three outputs not a judgment on the existence of a single fault, but a continuous chain of results from fault injection, protection action, topology switching, success or failure of automatic backup transfer, to the accumulation of heavy overload. As a parallel extension scheme, the fast-process solution can employ either implicit trapezoidal integration or discrete state transitions under piecewise constant input; the automatic backup transfer condition verification can use either a hard logic chain or a state machine corresponding to the plant configuration file. Neither path alters the set of fault scenarios. First round action sequence Predicting topology and heavy overload accumulation The output interface.

[0066] As a supplement: the data processing server reads the protection setting zone number, protection pressure plate status, CT / PT ratio, trip output binding relationship, and circuit breaker status; after fast process solution, the voltage and current phasors seen at the protection installation point are generated first; then, the distance protection zone, current protection start, zero-sequence protection start, and bus differential protection criteria are determined according to the protection type; when the protection start signal continuously reaches the set delay and the pressure plate is engaged, a trip command is generated; if the corresponding circuit breaker position does not change within the failure criterion time limit after tripping, a failure protection action sequence is generated.

[0067] When in use, a set of fault scenarios The admittance disturbance is written in first, and then the transient solution is entered. Therefore, the generation of protection actions is based directly on the fault electrical response rather than empirical enumeration. The first round of action sequence It has the same triggering sequence as the field protection circuit. Predictive topology. Simultaneously incorporating the results of protection cutoff, virtual AVC reallocation, and backup automatic transfer verification, the heavy overload accumulation amount obtained from slow process power flow recalculation was also included. With a clear physical source, it can be directly used for ring network differential pressure calculation in step four and for load reduction statistics and event classification in step five. Furthermore, step three puts the primary fault, secondary action, and topology consequences into the same scenario sandbox for continuous calculation, reducing the inference gaps caused by using different topology and state calibers in the preceding and following steps.

[0068] Step 4: Predicting Topology The above steps complete the tracing of power supply sources, calculation of ring network voltage difference, and solidification of fine-grained risk labels, thus establishing the predicted topology. The power supply source relationship is finely separated, and the voltage state and phase angle state at both ends of the tie switch are compared under the same criterion, so that structural risk and operational risk are closed in the same processing chain.

[0069] Step 3 has yielded the predicted topology after the fault. However, based solely on disconnected lines, de-energized busbars, and the shifted power flow, it is not possible to directly determine whether the network is in a vulnerable state where it appears to have dual power sources but is actually powered from the same source. Especially in 110kV and below distribution networks and regional ring networks, being energized in both directions does not mean that the two directions are independent; when both sides are ultimately traced back to the same upstream busbar, the same main transformer, or the same tower channel, performing loop closing, load switching, or interconnection power supply on site will still create an undesirable circulating current impact the moment the interconnection switch is closed.

[0070] The following actions are executed collaboratively by the data processing server, the simulation computing cluster, and the front-end 3D geographic interactive interface. The data processing server first reads the predicted topology output from step three. Node state vector Associated state vector and heavy overload accumulation Extract currently energized power supply points, busbar segments, main transformer branches, tie switches, and line corridor objects; then trace back the power supply chain layer by layer according to the feeder layer, substation layer, and upper-level power supply layer to generate a risk feature vector. The simulation computing cluster then performs virtual disconnection and recalculation of the voltage at both ends for each candidate tie switch without changing the rest of the network environment; the front-end 3D geographic interactive interface receives the risk feature vector. After determining the differential pressure result, different styles of risk signs are superimposed on the locations of substations, lines, and tie switches. Thus, step four no longer stops at the general conclusion that there is a risk, but instead specifies the source of the risk, the boundary of its effect, and the objects involved.

[0071] First, the data processing server is based on the predicted topology. Establish a hierarchical adjacency list, in which each connection edge carries the switch status, busbar affiliation, voltage level, and tower channel identifier.

[0072] Subsequently, starting from both ends of each load busbar or the tie switch to be operated, the process traces back along the currently energized connection side until it reaches the upstream power source point, generator grid connection point, or main grid receiving busbar. If only one valid power source point can be traced on one side, a single power source risk label is written; if both sides can be traced back separately but ultimately fall to the same upstream power source object, a same upstream power source risk label is written; if two power supply channels are topologically separate but their line corridor identifiers point to the same tower channel, a double-circuit risk label on the same tower is written. Its risk characteristic vector can be represented as: Where: Risk feature vector The fine-grained risk label set output in step four includes at least single-power-source risk tags, risk tags for the same upstream power source, risk tags for dual-circuit systems on the same pole, identification tags for substations, identification tags for lines, and identification tags for tie switches; mapping operator. The risk induction process performed by the data processing server takes the following form: first, based on the predicted topology... Construct charged connected components, and then utilize the associated state vectors. Extract line corridor, bay assignment, and bus attributes, and finally base the analysis on heavy overload aggregation. Sorting of risk labels involving channels; mapping operator Implementation process: First, predict the topology Delete the power-off edge and keep only the energized connected edge; then perform a breadth-first search starting from the busbars at both ends of the tie switch or the load busbar to trace its upstream power source; if the number of reachable power sources is one, mark it as a single power source risk; if the two power supply paths eventually trace back to the same upstream busbar or the same main transformer, mark it as the same upstream power source risk; if the critical lines on the two paths correspond to the same tower channel number, mark it as a double-circuit risk on the same tower.

[0073] Predicting topology Step 3 outputs the post-fault wiring relationships; accompanying state vectors. Steps one, two, and three continuously inherit the set of object attributes used to supplement tower channels, bay affiliations, and equipment types; heavy overload accumulation amount. The channel over-limit aggregation results output in step three are used to identify which risk labels should be prioritized for submission to the front-end interface. As a supplement: the line corridor number, tower channel number or GIS tower resource identifier required for identifying dual circuits on the same pole comes from the PMS ledger or GIS line resource database; in step four, when tracing back the power supply source, the line corridor object identifier is read at the same time; if two independent power supply channels correspond to the same tower channel identifier in the critical section, the dual circuit risk label on the same pole is written.

[0074] In one implementation, within a county's 110kV network, two substations supply power to the same industrial park from different directions, seemingly creating a dual-sided power supply. However, after tracing upstream along both busbars, the data processing server discovers that both power paths ultimately converge into the same section of the same busbar at a 220kV substation. Therefore, the relevant feeders to the industrial park are marked with the same upstream power risk flag. The front-end 3D geographic interface then displays the same source identifier next to the two incoming lines, showing maintenance personnel two lines with power but originating from the same source, rather than simply two green lines.

[0075] In risk feature vector After generation, the data processing server filters out a set of candidate tie switches with tie operation attributes and temporarily sets each candidate tie switch to a virtual disconnect state. The remaining lines, buses, main transformers, switches, and reactive power resources maintain the predicted topology obtained in step three. constant.

[0076] The simulation computing cluster then recalculated the voltage magnitude and phase angle of the buses at both ends of the tie switch, and converted the state at both ends into a dynamic voltage difference. If the dynamic pressure difference Exceeding the allowable differential pressure threshold If the contact switch is closed in the current network, it will be determined that this will introduce an over-limit circulating current risk; if the dynamic differential pressure is... Not exceeding the allowable differential pressure threshold If the condition is met, the contact switch is deemed to have met the conditions for loop closure. The criterion is written as: Where: dynamic pressure difference The equivalent voltage difference between the two busbars of the candidate tie switch in a virtual open state reflects the potential voltage inconsistency if a loop-closing operation is performed; bus voltage amplitude. : The busbar on one side of the tie switch at the time Voltage amplitude; bus voltage amplitude The busbar on the other side of the interconnecting switch is at the time Voltage amplitude; bus phase angle : The busbar on one side of the tie switch at the time The voltage phase angle; bus phase angle The busbar on the other side of the interconnecting switch is at the time Voltage phase angle; imaginary unit Imaginary unit in phasor operations; determination ratio Dynamic differential pressure Relative to the allowable differential pressure threshold The normalization result; when the judgment ratio When the value is greater than 1, write an out-of-limit circulation risk flag; Permissible differential pressure threshold The allowable differential pressure threshold is defined by a pre-configured upper limit of differential pressure based on voltage level, interconnection switch capacity, and dispatching operation procedures. The thresholds are derived from a table pre-set by voltage level, interconnection switch capacity, and local dispatching operating procedures; if not directly given by local procedures, they are generated by multiplying the rated voltage by the percentage of the allowable differential voltage.

[0077] In one implementation, a tie switch connects two 10kV busbars, and the site is preparing to change from a closed-loop to a closed-loop configuration. The data processing server first sets the tie switch to open in the twin space, and the simulation computing cluster then recalculates the busbar voltages on both sides of the switch. In the front-end 3D geographic interactive interface, the busbars on both sides of the switch maintain their respective energized colors, but a differential voltage indicator appears next to the switch body; if the differential voltage indicator exceeds a threshold, a warning color block is displayed around the switch, and maintenance personnel suspend the closed-loop operation accordingly.

[0078] As a parallel expansion scheme, the candidate tie switch can come from either the area where a power outage has already occurred in step three, or from a ring network area that is currently not faulty but has a shared power supply tag; both scenarios share the same dynamic differential pressure. And the judgment ratio This ensures interface consistency.

[0079] When used, predict the topology. The originally dispersed power supply chain is compressed into a risk feature vector. Therefore, the risks associated with single power supply, the same upstream power supply, and double circuits on the same pole can be attributed to specific substations, lines, and tie switches. The operational prerequisites for candidate tie switches no longer rely on experience-based judgment, but rather on the dynamic differential pressure after a virtual disconnection. And the judgment ratio The risk boundaries of the large-scale network closure and disclosure are given directly, thus allowing them to be revealed before execution.

[0080] Step 5: Convert the fault simulation results into executable event levels, power reduction details, and early warning outputs. Place the pre-fault baseline and post-fault baseline into the same comparison framework, first solidify the absolute power reduction load value, then separate the load transfer amount, then force the event level to match using a rule tree link, and finally write the results into a "Power Grid Risk Early Warning Notification" and push it to the front-end interface.

[0081] Steps three and four have already provided the predicted topology after the fault. Heavy overload accumulation Risk feature vector And the differential pressure judgment results of the interconnection switch, but these results are still scattered across different objects and different scenarios. When dealing with risks, dispatchers are more concerned with how much load reduction the scenario will cause, which important users will lose power, which loads will be taken over by other channels, and what event level these consequences correspond to.

[0082] The following actions are executed sequentially by the data processing server, rule tree engine, notification generator, and front-end 3D geographic interactive interface. The data processing server first reads the normal operation baseline status output in step two and the post-fault status output in step three, and establishes a load comparison list according to four levels: substation, busbar, feeder, and user access point. The rule tree engine then combines this with risk feature vectors... The system matches the automatic transfer switch action results, the list of important users, and the details of power reduction with the execution level. The notification generator then assembles the classification results, the power outage objects, the power transfer objects, and related line information into a notification text. The front-end 3D geographic interactive interface completes color overlay and pulse animation based on the object index in the notification text. The power outage status of important users does not change the thresholds for levels five, six, and seven; it is only used as a priority description item in the warning notification.

[0083] The data processing server does not treat all load changes as a reduction in power supply. Instead, it first distinguishes between three states: complete power outage, continued power supply after being transferred via a relay or backup channel, and power supply path change even though there is no power outage. To this end, the data processing server uses the normal operating baseline state in step two as the pre-fault reference and the post-fault state after completing the full sequence of actions in step three as the post-fault reference, comparing the active load carrying capacity of each user access point, each feeder, and each bus section one by one.

[0084] If an object is energized and carrying load before the fault, and is in a state of undervoltage or isolation after the fault, its corresponding load is included in the absolute load reduction value. If an object remains energized before and after the fault, but the power supply source is switched from the original channel to a backup channel or a connecting channel, its corresponding load is included in the load transfer amount but not in the absolute load reduction value. The quantification relationship is written as follows: Where: absolute load reduction value :time The total load of user access points and feeders that transition from energized to de-energized in fault scenarios is used to enter the event grading chain; load transfer amount. :time The total load that has not lost power but whose power supply path has switched during a fault scenario, used to reflect the pressure of power transfer and the availability of backup channels; user object set. The set of user access points, feeder ends, and important load objects included in this statistic directly inherits their object identifiers from the associated state vector. The smallest unit of measurement is the user access point; feeders and buses are only used for positioning and are not summed repeatedly. Pre-failure load : object Active load value in the normal operating baseline state in step two; power failure detection quantity The value can be 0 or 1; when the object Take 1 when there is a loss of voltage, isolation, or disconnection from the power supply chain due to protection action after a fault; otherwise, take 0. The determination is based on whether the system was energized and had a load before the fault, or whether there was a loss of voltage or isolation after the fault. (Power transfer discrimination quantity) The value can be 0 or 1; when the object The value is 1 if the power supply path identifier changes before and after the fault but the power supply is energized before and after the fault; otherwise, the value is 0, determined by the change in the power supply source number when the power supply is energized before and after the fault. In one implementation, after a section of the busbar in a 110kV substation loses power, two industrial feeders are energized via tie switches, while the remaining residential feeder remains de-energized due to a downstream switch malfunction. The data processing server compares the energized state and power supply path of the three feeders before and after the fault, and records the original load of the first two feeders as load transfer data. The original load of the next feeder is written into the absolute load reduction value. The notification generator will not mistakenly write industrial feeders that have had their power restored as objects with reduced power supply when writing subsequent notifications.

[0085] Preferably, the data processing server simultaneously reads the automatic transfer switch action result and the tie switch closing position result when comparing the object status; if the object experiences a short-term voltage loss but regains power before the end of the full-sequence action, it is still treated as a transferred object and is not included in the absolute load reduction value. As a parallel expansion solution, for critical users such as hospitals, rail transit stations, and data centers, it can be applied to user object sets. Add an importance tag to it so that it can get higher priority in the rule tree engine later.

[0086] In absolute load reduction values and load transfer amount Once formed, the rule tree engine matches data layer by layer in the following order: reduced load, busbar outage, critical user status, automatic transfer switch result, and risk label. The rule tree engine has three preset thresholds: absolute reduced load value. A load reduction of no less than 100MW, or a complete shutdown of any 220kV busbar, corresponds to a Level 5 event; absolute load reduction value. No less than 40MW, or a complete shutdown of the 110kV busbar, corresponding to a Level 6 event; absolute load reduction value. No less than 10MW, corresponding to a Level 7 event.

[0087] Then, important user power failure identifiers and risk feature vectors are superimposed. The risks associated with a single power source, the same upstream power source, and double return on the same pole, as well as the sequence of actions in the first round. The results of the backup self-cast action generate an event level quantity. Its rule compression expression is as follows: Where: event level quantity The event level result output by the rule tree engine has a value that is a discrete level identifier corresponding to a level 5, level 6, or level 7 event. Rule mapping operator The hierarchical matching process executed by the rule tree engine takes the following form: first, compare the absolute load reduction values. After checking the bus stop flag against the preset threshold, a risk feature vector is then introduced. and the first round of action sequence The backup self-transfer action results are used to perform level verification and detailed supplementation for scenarios under the same threshold; the first layer first judges... If any MW or 220kV busbar is completely shut down, a Level 5 event is output; otherwise, a judgment is made. If the MW or 110kV busbars are completely shut down, a Level 6 event is output; otherwise, a judgment is made. MW; if the condition is met, output a level 7 event; otherwise, output an event that does not meet the event level but retains the risk warning; the power failure status of important users is given priority in the notification field and the above level thresholds are not changed separately.

[0088] Absolute load reduction value Used as the primary axis for rating; risk feature vector Used to supplement the notification form with risk descriptions such as power supply from the same source, dual circuits on the same pole, and over-limit circulating current; predicting topology. From the first round of action sequence The automatic switching action results and equipment activation / deactivation status are jointly rewritten; the first round of action sequence. Used to extract information on the success of protection actions and automatic transfer switches, thereby determining whether the power outage is still ongoing; first-round action sequence. It is generated jointly by the protection device activation criteria, the delay criteria, the pressure plate status, and the circuit breaker position change results; Among them, busbars and feeders are only used for positioning and do not directly participate in the totaling; absolute load reduction values and load transfer amount All statistics are based solely on the set of user access points.

[0089] In one implementation, after receiving a fault scenario, the rule tree engine first identifies the absolute load reduction value. The preset threshold has been reached, and a section of 110kV busbar is completely shut down, triggering a Level 6 event flag. The notification generator then writes the level, the name of the power-destroyed feeder, the name of the residential area not yet restored, a description of the automatic transfer switch failure, a risk description of the same upstream power source, and the relevant tie switch number into the "Power Grid Risk Warning Notification." Subsequently, the front-end 3D geographic interactive interface, based on the object index in the notification, renders the outer contours of the relevant substations in a warning color, displays the power-destroyed line as a flashing pulse, and displays the transfer lines in a different emphasis style. On-duty personnel can directly distinguish between the power-destroyed objects and the objects currently receiving power transfer on the screen.

[0090] As a parallel extension scheme, the notification generator can output both document formats for manual review and structured messages for the dispatch master station to receive; both formats share the same event level. Absolute load reduction values and load transfer amount Therefore, the interaction method of the front-end interface in step five will not be changed.

[0091] When using, the absolute load reduction value With load transfer amount By separating and statistically analyzing data, objects with restored power supply will not be confused with those that are still experiencing power outages. Therefore, the classification focus in step five is more closely aligned with the final power supply status after the fault. The rule tree engine integrates power reduction thresholds, bus outages, important user status, backup automatic transfer action results, and risk feature vectors. The number of event levels generated when placed in the same decision chain It has a clear source and can directly drive notifications and front-end interfaces.

[0092] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0093] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0094] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0095] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0096] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A power grid risk early warning method based on digital twins, characterized in that: include, Acquire the CIME model and real-time cross-sectional data of the power grid energy management system (EMS), the equipment ledger of the production management system (PMS), and the real-time secondary protection data of the information protection system (IPS). Construct a physical model of the primary system, a mechanism model of the secondary system, and a digital twin base for associated operation modes. Calculate the baseline situation based on the digital twin base and virtual AVC constraints. Inject hypothetical equipment faults into the digital twin base, perform transient analysis, relay protection action prediction, backup automatic transfer action prediction, and power flow recalculation to obtain the predicted topology. Based on the predictive topology, single power source risk, same upstream power source risk and same pole double circuit risk are identified, and the dynamic pressure difference between the two ends is calculated after the virtual disconnection of the tie switch. The absolute load reduction and load transfer amount are quantified based on the load distribution before and after the fault, and an event level and power grid risk warning notification are generated accordingly.

2. The power grid risk early warning method based on digital twins according to claim 1, characterized in that: When constructing the digital twin base for the co-operation mode, object association, parameter identification and data cleaning are performed on the CIME model, equipment ledger and real-time data of secondary protection, and a unified object identifier is generated after establishing the correspondence between primary equipment and secondary devices.

3. The power grid risk early warning method based on digital twins according to claim 2, characterized in that: The accompanying operation mode digital twin base also includes an operation information model constructed based on real-time cross-sectional data and historical trends. The operation information model uniformly writes the current topology status, measurement status, and equipment commissioning / decommissioning status, and serves as the input cross-section for subsequent baseline situation calculations.

4. The power grid risk early warning method based on digital twins according to claim 3, characterized in that: The baseline situation for virtual AVC constraint calculation is taken into account, including constructing the admittance matrix based on the bus, line, main transformer, parallel reactive power equipment and switch status in the primary system physical model, and simultaneously taking into account the main transformer tap changer action sequence and reactive power compensation resource commissioning and decommissioning constraints during the power flow iteration process.

5. The power grid risk early warning method based on digital twins according to claim 4, characterized in that: After injecting hypothetical equipment faults into the digital twin base, a fast-process electromechanical transient analysis is first performed, and the first round of relay protection actions and corresponding isolation ranges are predicted based on the protection setting zone, pressure plate status, output circuit and circuit breaker status in the secondary system mechanism model.

6. The power grid risk early warning method based on digital twins according to claim 5, characterized in that: After the first round of relay protection actions causes topology changes, the process of predicting topology formation further includes: calling virtual AVC constraints to update the voltage distribution of the entire network, predicting automatic transfer switching actions based on the power failure side status, backup power supply side status and backup switch status, and performing power flow recalculation based on the action results.

7. The power grid risk early warning method based on digital twins according to claim 6, characterized in that: When identifying single-power-source risks, risks from the same upstream power source, and risks of dual circuits on the same pole, the power supply sources of the busbars and feeders are traced back along the energized connection relationships in the predicted topology, and the corresponding risk types are determined by combining the line corridor information and pole channel information in the equipment ledger.

8. The power grid risk early warning method based on digital twins according to claim 7, characterized in that: When calculating the dynamic voltage difference across the tie switch, while keeping the rest of the network environment in the predicted topology unchanged, the tie switch is virtually disconnected in the digital space, the voltage state of the bus at both ends of the tie switch is recalculated, and the dynamic voltage difference is compared with a preset threshold.

9. The power grid risk early warning method based on digital twins according to claim 8, characterized in that: When quantifying the absolute load reduction and load transfer, the load distribution status before the hypothetical equipment failure and the load distribution status after the completion of the full sequence operation are recorded. The user access point is used as the statistical object to distinguish between continuous power loss load and transferred load, and the power loss status of important users is extracted.

10. The power grid risk early warning method based on digital twins according to claim 9, characterized in that: When generating event levels and power grid risk warning notices, the absolute load reduction value, power outage status of important users, automatic transfer actions of backup power supplies, and details of load reduction are sent to the evaluation rule tree engine. Based on the load reduction threshold and the bus full shutdown condition, the event is determined to be a level 5, level 6, or level 7 event. The risky substations and lines are marked through the front-end dynamic three-dimensional geographic interactive interface linked by the RPC framework.

Images