Power grid fault deduction method and device based on scene evolution, equipment and medium

Abstract

The invention discloses a power grid fault deduction method and device based on scene evolution, equipment and a medium, and the method comprises the steps: constructing a risk evolution scene library of a target power grid according to the pre-obtained multi-source heterogeneous data of the target power grid; wherein the multi-source heterogeneous data comprises historical accident data, equipment parameter data, power grid operation data and external environment data; the risk evolution scene library comprises a plurality of scene triples; the scene triad is composed of states, reasons and judgment; mapping the risk evolution scene into a corresponding Petri network dynamic model; and calculating and setting a trigger probability for each transition in the Petri network dynamic model, and performing random cascading failure evolution simulation on the Petri network dynamic model for a plurality of times to obtain a cascading failure deduction result of the target power grid. According to the method, cascading failure dynamic deduction can be carried out on brand new risk scenes which are lack of historical data support or are caused by multi-factor complex coupling, and the generalization ability of power grid cascading failure deduction is improved.

Description

Technical Field

[0001] This invention relates to the field of power grid operation and maintenance management technology, and in particular to a method, apparatus, equipment and medium for power grid fault prediction based on scenario evolution. Background Technology

[0002] With the large-scale grid connection of new energy sources such as wind power and photovoltaics, and the widespread application of DC transmission and power electronic equipment in the power grid, the structure and operating characteristics of power systems are becoming increasingly complex, significantly increasing the risk of cascading failures. Cascading failures are typically triggered by initial disturbances and propagate tier by tier through multiple mechanisms such as electrical coupling, protection actions, and information exchange, potentially leading to severe consequences such as widespread power outages. To enhance the power grid's security and defense capabilities and achieve early warning and precise control of failures, conducting power grid failure simulations is crucial. Fault simulations aim to simulate the entire process of a failure from its occurrence and development to its final consequences, revealing its evolution path and key links, thereby providing decision-making support for operators, optimizing emergency plans, and reducing system operational risks.

[0003] Currently, there is considerable research in the field of power grid cascading failure simulation and risk assessment. Existing methods mainly rely on historical operating data, failure records, or simulation cases, using techniques such as statistical learning, causal inference, or attack graph modeling to construct failure propagation models and conduct risk analysis. For example, some methods, based on a large number of historical cascading failure scenarios, use machine learning models such as logistic regression and neural networks to extract the causal relationship between the initial failure and subsequent failures, and combine time-series features to sample and generate failure scenarios, or model the propagation paths of each failure, ultimately conducting risk assessment. However, these methods essentially rely on the statistical regularities and causal relationships inherent in historical data or known attack patterns, and their extrapolation capabilities are limited by the scope of historical experience. When the power grid faces new risk scenarios that exceed historical records—such as unprecedented extreme weather combinations, new equipment failure modes, or cascading effects caused by complex coupling of multiple factors—existing methods, lacking corresponding historical data support, struggle to effectively and reliably perform dynamic simulations and root cause analysis of these new scenarios, thus limiting their early warning and decision support capabilities in dealing with unknown and uncertain risks. Summary of the Invention

[0004] This invention provides a method, apparatus, equipment, and medium for power grid fault prediction based on scenario evolution. It can perform dynamic prediction of cascading faults in new risk scenarios, even those lacking historical data support or caused by complex coupling of multiple factors, through scenario evolution and dynamic simulation, thereby improving the generalization capability of power grid cascading fault prediction.

[0005] In a first aspect, embodiments of the present invention provide a power grid fault prediction method based on scenario evolution, comprising: Based on the pre-acquired multi-source heterogeneous data of the target power grid, a risk evolution scenario library for the target power grid is constructed. The multi-source heterogeneous data includes historical accident data, equipment parameter data, power grid operation data, and external environment data. The risk evolution scenario library includes several scenario triplets. Each scenario triplet consists of a state, a cause, and a judgment. The state represents the operating state or fault condition of the target power grid; the cause is the factor that causes the corresponding state to change; and the judgment is the conditional branch that causes the corresponding state to change. The risk evolution scenario library is mapped to a corresponding Petri net dynamic model; wherein, the Petri net dynamic model includes several locations, transitions, and directed arcs; the locations are obtained by mapping the states in the scenario triples; the transitions are obtained by mapping the judgments in the scenario triples and setting trigger conditions or attribute weights according to the causes in the scenario triples; the directed arcs are used to connect the locations and transitions according to preset logical associations; For each transition in the Petri net dynamic model, a trigger probability is calculated and set, and several random cascading failure evolution simulations are performed on the Petri net dynamic model with the trigger probability to obtain the cascading failure inference results of the target power grid; wherein, the trigger probability is calculated based on physical laws or historical decision data.

[0006] This invention provides a data and knowledge foundation for dynamic simulation and handling of new scenarios by transforming unstructured, multi-source, heterogeneous raw data (history, equipment, operation, environment) into structured, computer-processable knowledge units with clear causal logic (state-cause-judgment). This allows simulations to move beyond relying on specific historical cases or the causal logic of single faults and instead be based on general state transition logic. By converting the scenario knowledge base into a dynamic system model and accurately describing concurrent, asynchronous, and random behaviors, it provides a precise, computable model framework for subsequent large-scale, automated, and repeatable simulations. By assigning quantified triggering probabilities to each uncertain link (transition) in the model and exploring all possible evolutionary branches through random simulation, the probability distribution of various evolutionary final states and specific evolutionary paths are obtained through simulation, thereby improving the comprehensiveness of the simulation exploration of each cascading fault. Compared with existing technologies that model relatively static fault propagation chains based on power grid operation principles or integrate and analyze historical fault data, this invention can dynamically extrapolate the entire process of cascading faults in new risk scenarios, even those lacking historical data support or caused by complex coupling of multiple factors, through scenario evolution and dynamic stochastic simulation. This improves the generalization ability and risk tracing ability of power grid cascading fault extrapolation.

[0007] In some preferred embodiments of the first aspect, the step of constructing a risk evolution scenario library for the target power grid based on pre-acquired multi-source heterogeneous data of the target power grid specifically includes: Natural language processing technology is used to extract key accident data from the historical accident data to form a structured historical event record; wherein, the key accident data includes power grid failure events, failure causes and maintenance operations; Based on the preset power grid operation procedures, and using the equipment parameter data and power grid operation data, power grid operation logic analysis is performed to obtain the operation rules of the target power grid. By using a preset association rule mining algorithm, the temporal correlation between the power grid fault events is analyzed based on the equipment parameter data, power grid operation data, and external environment data, so as to obtain the event association rules of the target power grid. Based on the historical event records and operating rules, several scenario triplets are constructed, and each scenario triplet is logically connected according to the event association rules to construct the risk evolution scenario library of the target power grid.

[0008] This invention employs a comprehensive, multi-faceted intelligent knowledge mining method to integrate complex and disorganized power grid data or operating principles into triplets representing the formation principles of different system states or risk scenarios. It also mines the operational logic relationships between various risk scenarios, providing a structured and comprehensive knowledge base for subsequent dynamic model construction. This reduces the complexity of model construction while ensuring the accuracy of subsequent simulations.

[0009] In some preferred embodiments of the first aspect, the risk evolution scenario library is mapped to a corresponding Petri net dynamic model, specifically as follows: Map the state in each of the scenario triples to a place; The judgments in each scenario triple are mapped to one or more transitions, and trigger conditions or attribute weights are set for the transitions based on the reasons in the corresponding scenario triples. Based on the event association rules, the logical associations between the scenario triples in the risk evolution scenario library are determined, and based on the logical associations, the libraries and transitions are connected by directed arcs to construct an initial Petri net dynamic model. The initial Petri net dynamic model is subjected to reachability verification to obtain the final Petri net dynamic model.

[0010] This invention, through its embodiments, precisely maps each triplet element in the risk evolution scenario library to each component of the Petri net and introduces model verification methods, thereby accurately transforming the scenario logic into a dynamic model. This ensures that subsequent simulations strictly and consistently follow the operating principles of the target network, improving the accuracy of the simulation results.

[0011] In some preferred embodiments of the first aspect, calculating and setting a trigger probability for each transition in the Petri net dynamic model includes: Using a pre-acquired set of objective parameter sources, each transition in the Petri net dynamic model is sequentially matched, and the matched objective data source is determined as the trigger probability of the corresponding transition. The set of objective parameter sources includes the success rate of historical protection actions, the probability of equipment failure, and the probability of coupling risks. The matching operation includes keyword matching and type mapping.

[0012] This invention, by prioritizing the use of objective parameter sources (such as historical action success rate and equipment failure probability) to assign values ​​to the transition trigger probability, maximizes the use of objective data and physical laws, reduces the subjective arbitrariness of state transition triggering during dynamic evolution, and enhances the scientific rigor and credibility of evolution process simulation.

[0013] In some preferred embodiments of the first aspect, after determining the objective data source for matching as the corresponding trigger probability of the transition, the following are included: For transitions that are not matched with objective data sources, a cloud model is generated and associated with each transition based on pre-acquired historical decision data using a preset knowledge modeling algorithm; wherein, the cloud model includes the expected trigger probability, trigger probability entropy, and trigger probability hyper-entropy of the transition.

[0014] This invention introduces a cloud model to transform historical decision-making data into a digital model with probabilistic and statistical characteristics, enabling expert knowledge to be integrated with objective data within the same framework, ensuring that the model can still operate reasonably even when information is incomplete.

[0015] In some preferred embodiments of the first aspect, the Petri net dynamic model with triggering probabilities is subjected to several random cascading failure evolution simulations to obtain the cascading failure deduction results of the target power grid, specifically as follows: Using the Monte Carlo simulation method, several random cascading failure evolution simulations were performed on the Petri net dynamic model with triggering probabilities, and the state transition path and corresponding final system state of each random cascading failure evolution simulation were obtained to obtain the failure evolution simulation log. Based on the fault evolution simulation log, risk analysis is performed on each state transition path, the final system state, and the transitions in the state transition path to obtain the risk analysis results. By integrating the fault evolution simulation logs and the corresponding risk analysis results, the cascading fault projection results of the target power grid are obtained.

[0016] This invention employs Monte Carlo simulation to perform several evolutionary simulations and conducts cause analysis on each node of the evolutionary simulation results to provide complete and comprehensive evolutionary simulation results, providing comprehensive and accurate data for fault repair decisions and root cause tracing.

[0017] In some preferred embodiments of the first aspect, the Petri net dynamic model with triggering probabilities is subjected to several random cascading failure evolution simulations using the Monte Carlo simulation method. Specifically, the process of performing a single cascading failure evolution simulation is as follows: The Petri net dynamic model is initialized based on the pre-stored initial fault setting data to determine the initial token distribution for each location. The algorithm iterates through each enableable transition, using real-time generated random numbers and their corresponding trigger probabilities to determine whether each enableable transition is triggered in the current simulation. For transitions that are determined to be triggered, the token distribution in the input and output places of the transition is updated to update the state of the Petri net dynamic model, until no enableable transitions remain. Here, an enableable transition refers to a transition in which all input places have an initial token; an input place refers to a place located at the source of the corresponding directed arc; and an output place refers to a place located at the end of the corresponding directed arc. The final token distribution is determined as the final system state of this simulation, and the token transfer path in this simulation is determined as the state transition path of this simulation.

[0018] This invention utilizes Petri nets with trigger probabilities to perform dynamic evolution simulation, simulating various possible evolution paths and selecting evolution paths based on certain trigger probabilities, thus objectively and scientifically presenting the entire process of fault risk from its inception to its outbreak.

[0019] Secondly, embodiments of the present invention provide a power grid fault simulation device based on scenario evolution, comprising a scenario database construction module, a dynamic model construction module, and a power grid fault simulation module, wherein... The scenario library construction module is used to construct a risk evolution scenario library for the target power grid based on pre-acquired multi-source heterogeneous data of the target power grid. The multi-source heterogeneous data includes historical accident data, equipment parameter data, power grid operation data, and external environment data. The risk evolution scenario library includes several scenario triplets. Each scenario triplet consists of a state, a cause, and a judgment. The state represents the operating state or fault condition of the target power grid; the cause is the factor that causes the corresponding state to change; and the judgment is the conditional branch that causes the corresponding state to change. A dynamic model construction module is used to map the risk evolution scenario library to a corresponding Petri net dynamic model; wherein, the Petri net dynamic model includes several locations, transitions, and directed arcs; the locations are obtained by mapping the states in the scenario triples; the transitions are obtained by mapping the judgments in the scenario triples and setting trigger conditions or attribute weights according to the causes in the scenario triples; the directed arcs are used to connect the locations and transitions according to preset logical associations; The power grid fault simulation module is used to calculate and set the trigger probability for each transition in the Petri net dynamic model, and to perform several random cascading fault evolution simulations on the Petri net dynamic model with the trigger probability to obtain the cascading fault simulation results of the target power grid; wherein the trigger probability is calculated based on physical laws or historical decision data.

[0020] This invention, through a scenario database construction module, transforms unstructured, multi-source, heterogeneous raw data (history, equipment, operation, environment) into structured, computer-processable knowledge units with clear causal logic (state-cause-judgment). This provides the data and knowledge foundation for dynamic simulation and handling of new scenarios, enabling simulations to move beyond specific historical cases or the causal logic of single faults and instead rely on general state transition logic. The dynamic model construction module converts the scenario knowledge base into a dynamic system model, accurately describing concurrent, asynchronous, and stochastic behaviors, providing a precise, computable model framework for subsequent large-scale, automated, and repeatable simulations. The power grid fault simulation module assigns quantified triggering probabilities to each uncertain link (transition) in the model and explores all possible evolutionary branches through stochastic simulation. Simulation yields the probability distribution of various evolutionary final states and specific evolutionary paths, improving the comprehensiveness of cascading fault simulation exploration.

[0021] Thirdly, embodiments of the present invention provide a terminal device, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus; The memory is used to store at least one executable instruction that causes the processor to perform the operation of the scenario-based power grid fault simulation method as described above.

[0022] Fourthly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device or apparatus where the computer-readable storage medium is located to execute the power grid fault simulation method based on scenario evolution as described in any one of the above.

[0023] The above description is merely an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0024] Figure 1This is a schematic diagram of a power grid fault simulation method based on scenario evolution provided by an embodiment of the present invention; Figure 2 A flowchart illustrating the construction of a risk evolution scenario library, as exemplified by an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the conversion logic from the evolutionary relationship of scenario triples to a Petri net model, as exemplified by an embodiment of the present invention. Figure 4 A schematic diagram illustrating the execution flow of a single Monte Carlo simulation in a Petri net, as exemplified by an embodiment of the present invention; Figure 5 A schematic diagram illustrating a scenario-based power grid fault deduction method, as exemplified by an embodiment of the present invention; Figure 6 This is a structural diagram of a power grid fault simulation device based on scenario evolution, provided in an embodiment of the present invention. Detailed Implementation

[0025] 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.

[0026] Example 1: like Figure 1 As shown, an embodiment of the present invention provides a power grid fault prediction method based on scenario evolution, comprising the following steps: S101, Based on the pre-acquired multi-source heterogeneous data of the target power grid, construct a risk evolution scenario library for the target power grid; wherein, the multi-source heterogeneous data includes historical accident data, equipment parameter data, power grid operation data, and external environment data; the risk evolution scenario library includes several scenario triplets; each scenario triplet consists of a state, a cause, and a judgment; the state is the operating state or fault condition of the target power grid; the cause is the factor that causes the corresponding state to change; the judgment is the conditional branch that causes the corresponding state to change; In this embodiment, the step of constructing a risk evolution scenario library for the target power grid based on pre-acquired multi-source heterogeneous data of the target power grid specifically involves: extracting key accident data from the historical accident data using natural language processing technology to form a structured historical event record; wherein the key accident data includes power grid fault events, fault causes, and maintenance operations; performing power grid operation logic analysis based on the equipment parameter data and power grid operation data according to a preset power grid operation procedure to obtain the operation rules of the target power grid; analyzing the temporal correlation between the power grid fault events using a preset association rule mining algorithm based on the equipment parameter data, power grid operation data, and external environment data to obtain the event association rules of the target power grid; constructing several scenario triples based on the historical event records and operation rules, and logically connecting each scenario triple according to the event association rules to construct the risk evolution scenario library of the target power grid.

[0027] In one specific embodiment, the step of constructing a risk evolution scenario library for the target power grid based on pre-acquired multi-source heterogeneous data of the target power grid specifically involves: (1) Input data A. Historical Accident Reports (Text): Records past power grid failure events, their causes, and the handling process.

[0028] B. Equipment parameter library (structured data): contains physical parameters of equipment such as transmission lines, towers, and transformers (such as design wind speed, icing thickness, and service life).

[0029] C. Operation log and SCADA data (timing data): Records information such as the action sequence of protection devices, switch changes, and power flow exceeding limits.

[0030] D. External environmental data (spatiotemporal data): refined meteorological forecast data (wind speed, precipitation, temperature), geographic information system (GIS) data.

[0031] (2) Data processing and knowledge extraction operations Step 1 (Event Extraction): Using Natural Language Processing (NLP) technology, key information is automatically extracted from historical incident reports to form a structured event record.

[0032] The structured event record includes, for example: <time, device ID, event type (e.g., wind deflection flashover), external cause (e.g., wind speed 30m / s), and action taken>.

[0033] Step 2 (Association and Rule Mining): Perform association analysis on the time-series operation logs (e.g., using the Apriori algorithm) to uncover the chronological and causal relationships between equipment failures, protection actions, and power flow transitions. Simultaneously, encode the operation rules from documents such as the "Guidelines for Power System Safety and Stability" and the "Relay Protection Setting Procedures" into logical rules in the form of "IF-THEN".

[0034] Step 3 (SCJ Unit, i.e. Scenario Triple Generation): Based on the above output, SCJ units are automatically or assistedly generated.

[0035] 1) Status (S): The system status extracted from the event log, such as "500kV line L1 tripped".

[0036] 2) Cause (C): Combine external causes in the event log and internal states in the equipment parameter library (such as "equipment aging degree > threshold") to form a combined cause, such as "wind speed exceeds threshold & tower bolts are loose".

[0037] 3) Judgment (J, Judge): Derived from the coded operation rules or historical processing records, it defines the conditional branches for the system to evolve from one state to the next, such as "Was the reclosing successful?" The initial value of its branch probability can be derived from historical statistics (such as the reclosing success rate).

[0038] (3) Output A structured SCJ scenario knowledge base (i.e., risk evolution scenario base), in which each SCJ unit is a basic logical block that can be recognized and processed by a computer, together constitutes the prototype of a complete causal network for power grid risk evolution.

[0039] To better explain the specific working principles and data processing flow of the risk evolution scenario library construction, please refer to [link / reference]. Figure 2 , Figure 2 This is an example of a flowchart illustrating the construction of a risk evolution scenario library, demonstrating the entire process from multi-source heterogeneous raw data to structured SCJ knowledge units. The top layer represents four key data source inputs, the middle layer represents three core data processing and knowledge extraction technologies, and the bottom layer represents the output structured SCJ knowledge units. The entire process demonstrates how data is transformed into causal logical knowledge that can be processed by computers through specific technical means.

[0040] S102, the risk evolution scenario library is mapped to a corresponding Petri net dynamic model; wherein, the Petri net dynamic model includes several locations, transitions, and directed arcs; the locations are obtained by mapping the states in the scenario triples; the transitions are obtained by mapping the judgments in the scenario triples and setting trigger conditions or attribute weights according to the causes in the scenario triples; the directed arcs are used to connect locations and transitions according to preset logical associations; In this embodiment, the risk evolution scenario library is mapped to a corresponding Petri net dynamic model. Specifically, the state in each scenario triplet is mapped to a location; the judgment in each scenario triplet is mapped to one or more transitions, and trigger conditions or attribute weights are set for the transitions according to the causes in the corresponding scenario triplets; the logical association between each scenario triplet in the risk evolution scenario library is determined according to the event association rules, and each location and transition is connected by directed arcs according to the logical associations to construct an initial Petri net dynamic model; the initial Petri net dynamic model is subjected to reachability verification to obtain the final Petri net dynamic model.

[0041] It should be noted that, for the convenience of computer simulation and mathematical processing, the visualized scenario evolution diagram needs to be converted into a formal Petri net model. Petri nets consist of places (P), transitions (T), and directed arcs (Arc), and can accurately describe the behavior of concurrent, asynchronous, and stochastic systems.

[0042] In one specific embodiment, the risk evolution scenario library is mapped to a corresponding Petri net dynamic model, specifically as follows: (1) Input: Pre-generated structured SCJ knowledge base.

[0043] (2) Model synthesis operation: Step 1 (Element Mapping): 1) Map each state to a place (P). A place represents a possible state of the system, and the number of tokens in a place indicates the occurrence of that state.

[0044] 2) Map each judgment to one or more transitions (T). A judgment usually corresponds to two transitions, representing the "yes" (evolution towards improvement) and "no" (evolution towards decline) branches respectively.

[0045] 3) The cause information is incorporated into the triggering conditions or attribute weights of the transition.

[0046] Step 2 (Establishment of connecting arc): Based on the logical order of the SCJ units, directed arcs (Arc) connect the places and transitions. The direction of the arc indicates the direction of the state transition: from the upstream state (place) to the decision (transition), and then from the transition to the downstream state (place).

[0047] Step 3 (Model Integration and Validation) All locations, transitions, and arcs obtained from the mapping are integrated to form a complete Petri net model. Model validation tools are then used to check for fundamental properties such as deadlock and reachability.

[0048] (3) Output Through this transformation, a scenario evolution graph containing complex logic and branches is transformed into a well-defined Petri net model, whose mathematical representation is: PN=(P,T,F,M0) In the formula: A finite set of libraries, representing the various states that the system may be in; T A finite set of transitions, representing the judgments or events that lead to a state transition; F⊆(P×T)∪(T×P): A set of directed arcs, defining the flow relationship between states and events; : Initial identifier function, which defines the distribution of token quantity in each vault at the initial moment.

[0049] To better explain the specific working principle and data processing flow of the Petri net model construction, please refer to [link / reference]. Figure 3 , Figure 3 This is a schematic diagram illustrating the conversion logic from the evolutionary relationship of scenario triples to a Petri net model, as exemplified by an embodiment of the present invention.

[0050] S103, calculate and set the trigger probability for each transition in the Petri net dynamic model, and perform several random cascading failure evolution simulations on the Petri net dynamic model with the trigger probability to obtain the cascading failure inference results of the target power grid; wherein, the trigger probability is calculated based on physical laws or historical decision data.

[0051] In this embodiment, calculating and setting the trigger probability for each transition in the Petri net dynamic model includes: performing a matching operation on each transition in the Petri net dynamic model sequentially using a pre-acquired set of objective parameter sources, and determining the matching objective data source as the trigger probability of the corresponding transition; wherein, the set of objective parameter sources includes the success rate of historical protection actions, the probability of equipment failure, and the probability of coupling risk; the matching operation includes keyword matching and type mapping.

[0052] For example, in scenarios involving relay protection actions, the probability of "correct operation of the main protection" can be set based on historical correct operation rate statistics of the protection device of that model.

[0053] For example, in the case of equipment failure scenarios, the probability of "old towers collapsing in winds of level X" can be calculated using a mechanical model based on the tower's design parameters, the current wind speed, and structural health monitoring (SHM) data.

[0054] For example, in scenarios involving multiple factors, the NK model is used to analyze the coupling results of "people, objects, environment, and management" factors in the historical accident database (such as the highest risk of "object-environment" coupling). When modeling, the focus is on identifying and setting changes involving the combined effects of equipment and environment, and the probability of these changes must take into account both equipment status and environmental stress.

[0055] In this embodiment, after determining the matching objective data source as the trigger probability of the corresponding transition, the process includes: for transitions that do not match the objective data source, generating and associating each transition with its own cloud model based on pre-acquired historical decision data using a preset knowledge modeling algorithm; wherein, the cloud model includes the expected trigger probability, trigger probability entropy, and trigger probability hyper-entropy of the corresponding transition.

[0056] In one specific embodiment, for changes that do not match objective data sources, respective cloud models are generated, specifically as follows: For judgments that must rely on expert experience, or to characterize the uncertainty of objective parameters themselves (such as the error in predicting wind speed), cloud models are used for processing.

[0057] 1) Establish an evaluation benchmark cloud Mapping the set of probabilities linguistic values ​​V = {extremely unlikely, somewhat unlikely, moderately likely, somewhat likely, extremely likely} onto the universe of discourse U = [0,1], we generate five standard normal cloud models as benchmarks using the golden section method. Each cloud is characterized by three numerical features: Expectation (Ex): The central value of a concept in the universe of discourse; Entropy (En): The ambiguity and randomness of a concept; the greater the entropy, the more ambiguous the concept. Hyperentropy (He): A measure of the uncertainty of entropy.

[0058] Furthermore, the digital features of the scale cloud are calculated using the following formula: The formula for calculating the baseline cloud (which is likely to occur): Formula for calculating adjacent clouds (most likely to occur): in, , , Corresponding to the cloud model of "generally possible", , , This corresponds to the cloud model that is "more likely to occur". Other cloud models are generated symmetrically according to this pattern.

[0059] 2) Generate transition judgment cloud Domain expert groups refer to a collection of experts with profound theoretical knowledge and rich practical experience in specific professional fields (such as power system operation, equipment fault diagnosis, relay protection, etc.). In knowledge-based system modeling and simulation, this group plays a core role in providing experiential knowledge, quantifying subjective judgments, and modeling uncertainties, thereby enhancing the system's generalization ability and decision-making credibility when dealing with new and complex risk scenarios. In the method described in this invention, the introduction of domain expert groups is mainly used to address decision-making stages where data is scarce or cannot be fully objectively quantified (such as setting the probability of transition triggering). A domain expert group E={e1,e2,...,e_q} is invited to evaluate the probability of each transition Tᵢ occurring, providing their perceived minimum and maximum values. After collecting the evaluations from all experts, a comprehensive judgment cloud Cᵢ representing the uncertainty of the probability of the transition is generated using a reverse cloud generator algorithm. , , The core formula of the algorithm is as follows: expect : entropy : hyperentropy : In the formula: q: total number of experts , : The minimum and maximum estimates of the probability of change Tᵢ by the kth expert.

[0060] 3) The conditions for determining the activation rule of a transition Tᵢ are: in, These are sample values ​​randomly generated from the judgment cloud Cᵢ. It is the scale cloud interval corresponding to the language level Li determined by experts.

[0061] The final result is a parameter-complete, executable Petri net simulation model, in which each transition is associated with a specific trigger probability or trigger probability distribution (cloud model).

[0062] In this embodiment, several random cascading failure evolution simulations are performed on the Petri net dynamic model with trigger probabilities to obtain the cascading failure prediction results of the target power grid. Specifically, the Monte Carlo simulation method is used to perform several random cascading failure evolution simulations on the Petri net dynamic model with trigger probabilities, and the state transition path and corresponding final system state of each random cascading failure evolution simulation are obtained to obtain a failure evolution simulation log. Based on the failure evolution simulation log, risk analysis is performed on each state transition path, the final system state, and the changes in the state transition path to obtain risk analysis results. The failure evolution simulation log and the corresponding risk analysis results are integrated to obtain the cascading failure prediction results of the target power grid.

[0063] In this embodiment, the Petri net dynamic model with trigger probabilities is simulated several times using the Monte Carlo simulation method to simulate the evolution of random cascading failures. Specifically, the process of simulating a single cascading failure is as follows: The Petri net dynamic model is initialized based on pre-stored initial fault setting data to determine the initial token distribution for each place; each enableable transition is traversed, and using real-time generated random numbers and corresponding trigger probabilities, it is determined whether each enableable transition in this simulation is triggered. For transitions that are determined to be triggered, the token distribution in the input and output places of the transition is updated to update the state of the Petri net dynamic model until no enableable transitions remain; where an enableable transition refers to a transition where all input places have initial tokens; an input place refers to a place at the source of a corresponding directed arc; and an output place refers to a place at the end of a corresponding directed arc; the final token distribution is determined as the final system state of this simulation, and the token transition path in this simulation is determined as the state transition path of this simulation.

[0064] In one specific embodiment, the process of performing cascading failure evolution simulation is as follows: (1) Input: A Petri net model with complete parameters.

[0065] (2) Simulation operation A. Set the total number of simulations N (usually N ≥ 10000).

[0066] B. Single simulation run: 1) Initialize the network and set the token distribution M0 according to the initial fault; 2) Iterate through all transitions that enable (all input libraries have tokens); 3) For each enabling transition Ti, determine whether to trigger it based on its parameter weighting method: If the objective probability is p, then a uniformly distributed random number in the range [0,1] is generated; if the number is less than p, then a trigger is activated. If associated cloud model Ci, then determine whether to trigger according to the preset transition activation rules; 4) Trigger all eligible transitions, consume the input token, generate a new token in the output token, and update the network state M; 5) Repeat steps 2)-4) above until no transitions are enabled in this simulation. Record the final token distribution. The set of places holding tokens in Mend Mend represents the final evolutionary state reached in this simulation (e.g., "local recovery", "load loss 30%").

[0067] C. Multiple Simulations and Statistics: Run the independent simulation N times. Count the number of times each evolutionary final state Ek (corresponding to a specific location or group of locations) of interest occurs (Count(Ek)).

[0068] (3) Output: Empirical probabilities of each evolutionary final state: ; In the formula, For the end The estimated probability of occurrence For the final state in N simulations The probability of occurrence, where N is the total number of simulations, objectively reflects the likelihood of the system evolving in a specific direction under given uncertainty.

[0069] Complete simulation process log: Records the detailed path of token flow in each simulation, that is, the specific evolution path sequence.

[0070] To better illustrate the specific working principle and data processing flow of this invention's simulation, please refer to [link / reference]. Figure 4 , Figure 4 This is a schematic diagram illustrating the execution flow of a single Monte Carlo simulation in a Petri net, as exemplified by an embodiment of the present invention.

[0071] In one specific embodiment, the process of obtaining the final fault simulation result is as follows: (1) Input: Probabilities of each evolutionary final state And simulation process logs; Quantitative value of the severity of consequences S( Based on objective indicators such as load loss, power outage duration, economic loss, and social impact, or in conjunction with expert evaluation of indicator weights, calculate each final state. The severity score.

[0072] (2) Risk assessment and path backtracking operation: Step 1 (Risk Value Calculation and Rating): Calculate the risk value R for each final state. )= × S( Based on the preset risk matrix, R( This is mapped to a risk level (e.g., levels I-IV).

[0073] Step 2 (Critical Evolutionary Path Extraction): This is the core step in transforming probabilities into specific paths.

[0074] Path backtracking: For each high-risk (Level I, Level II) final state Extract all causes from simulation logs The evolution path is {Path1, Path2, ..., Pathm}.

[0075] Path probability allocation: Statistical analysis of the probability distribution of each path Pathj in the path leading to... The frequency of occurrence in the simulation is used to calculate its conditional probability P(Pathj| The contribution of this path to the total risk can be approximated as follows: ×P(Pathj| ).

[0076] Key node identification: In all high-risk paths, count the frequency of each transition (judgment point). The transitions that occur most frequently are the common key decision-making links or vulnerable links that lead to high-risk outcomes.

[0077] (3) Output (Decision Support) Risk heat map: On the power grid geographic map, the risk heat map is visualized based on the frequency and contribution of equipment in different high-risk paths.

[0078] List of key nodes and paths: List A (Key Intervention Points): A list of key transitions ordered by frequency of occurrence, such as "Transition T8 (Critical Path Power Flow Exceedance Judgment)", along with its objective parameters (such as the excess limit value) and current status.

[0079] List B (High-Risk Evolutionary Chain): List the top few specific evolutionary paths with the highest contribution, in the format: "Initial Failure → Result of Judgment 1 → Result of Judgment 2 → ... → High-Risk Final State", and indicate the contribution of each path.

[0080] Decision Priority Report: Combining List A and B, generate targeted action recommendations. For example: "For critical intervention point T8, it is recommended to immediately inspect and reinforce line L1 / L2; for high-frequency path P5, it is recommended to activate contingency plan V3 and adjust power generation output in advance." For a better explanation of the specific working principle and data processing flow of this invention, please refer to [link / reference]. Figure 5 , Figure 5 This is a schematic diagram illustrating a scenario-based power grid fault simulation method, as exemplified by an embodiment of the present invention.

[0081] This invention provides a data and knowledge foundation for dynamic simulation and handling of new scenarios by transforming unstructured, multi-source, heterogeneous raw data (history, equipment, operation, environment) into structured, computer-processable knowledge units with clear causal logic (state-cause-judgment). This allows simulations to move beyond relying on specific historical cases or the causal logic of single faults and instead be based on general state transition logic. By converting the scenario knowledge base into a dynamic system model and accurately describing concurrent, asynchronous, and random behaviors, it provides a precise, computable model framework for subsequent large-scale, automated, and repeatable simulations. By assigning quantified triggering probabilities to each uncertain link (transition) in the model and exploring all possible evolutionary branches through random simulation, the probability distribution of various evolutionary final states and specific evolutionary paths are obtained through simulation, thereby improving the comprehensiveness of the simulation exploration of each cascading fault. Compared with existing technologies that model relatively static fault propagation chains based on power grid operation principles or integrate and analyze historical fault data, this invention can dynamically extrapolate the entire process of cascading faults in new risk scenarios, even those lacking historical data support or caused by complex coupling of multiple factors, through scenario evolution and dynamic stochastic simulation. This improves the generalization ability and risk tracing ability of power grid cascading fault extrapolation.

[0082] Example 2: like Figure 6 As shown, this embodiment provides a power grid fault simulation device based on scenario evolution, including a scenario database construction module 201, a dynamic model construction module 202, and a power grid fault simulation module 203, wherein... The scenario library construction module 201 is used to construct a risk evolution scenario library for the target power grid based on the pre-acquired multi-source heterogeneous data of the target power grid; wherein, the multi-source heterogeneous data includes historical accident data, equipment parameter data, power grid operation data and external environment data; the risk evolution scenario library includes several scenario triples; the scenario triples consist of state, cause and judgment; In this embodiment, the scenario database construction module 201 constructs a risk evolution scenario database for the target power grid based on pre-acquired multi-source heterogeneous data of the target power grid. Specifically, the scenario database construction module 201 uses natural language processing technology to extract key accident data from the historical accident data to form a structured historical event record. The key accident data includes power grid fault events, fault causes, and maintenance operations. Based on the preset power grid operation procedures and the equipment parameter data and power grid operation data, a power grid operation logic analysis is performed to obtain the operation rules of the target power grid. Using a preset association rule mining algorithm, the temporal correlation between the power grid fault events is analyzed based on the equipment parameter data, power grid operation data, and external environment data to obtain the event association rules of the target power grid. Based on the historical event records and operating rules, several scenario triplets are constructed, and each scenario triplet is logically connected according to the event association rules to construct the risk evolution scenario library of the target power grid.

[0083] The dynamic model construction module 202 is used to map the risk evolution scenario library to a corresponding Petri net dynamic model; wherein, the Petri net dynamic model includes several locations, transitions, and directed arcs; the locations are obtained by mapping the states in the scenario triples; the transitions are obtained by mapping the judgments in the scenario triples and setting trigger conditions or attribute weights according to the causes in the scenario triples; the directed arcs are used to connect the locations and transitions according to preset logical associations; In this embodiment, the dynamic model construction module 202 maps the risk evolution scenario library to a corresponding Petri net dynamic model. Specifically, the dynamic model construction module 202 maps the state in each scenario triplet to a place; maps the judgment in each scenario triplet to one or more transitions, and sets trigger conditions or attribute weights for the transitions according to the causes in the corresponding scenario triplets; determines the logical association between each scenario triplet in the risk evolution scenario library according to the event association rules, and connects each place and transition through directed arcs according to the logical associations to construct an initial Petri net dynamic model; and performs reachability verification on the initial Petri net dynamic model to obtain the final Petri net dynamic model.

[0084] The power grid fault simulation module 203 is used to calculate and set the trigger probability for each transition in the Petri net dynamic model, and to perform several random cascading fault evolution simulations on the Petri net dynamic model with the trigger probability to obtain the cascading fault simulation results of the target power grid; wherein the trigger probability is calculated based on physical laws or historical decision data.

[0085] In this embodiment, the power grid fault simulation module 203 calculates and sets the trigger probability for each transition in the Petri net dynamic model, including: the power grid fault simulation module 203 performs a matching operation on each transition in the Petri net dynamic model in sequence through a pre-acquired set of objective parameter sources, and determines the trigger probability of the corresponding transition based on the matched objective data source; wherein, the set of objective parameter sources includes the historical protection action success rate, equipment failure probability, and coupling risk probability; the matching operation includes keyword matching and type mapping.

[0086] In this embodiment, after determining the matching objective data source as the trigger probability of the corresponding transition, the power grid fault inference module 203 includes: for transitions that do not match the objective data source, generating and associating their respective cloud models for each transition based on pre-acquired historical decision data using a preset knowledge modeling algorithm; wherein, the cloud model includes the expected trigger probability, trigger probability entropy, and trigger probability super-entropy of the corresponding transition.

[0087] In this embodiment, the power grid fault simulation module 203 uses the Monte Carlo simulation method to perform several random cascading fault evolution simulations on the Petri net dynamic model with trigger probabilities. Specifically, the process of performing a single cascading fault evolution simulation involves: the power grid fault simulation module 203 initializing the Petri net dynamic model based on pre-stored initial fault setting data, determining the initial token distribution for each location; traversing each enableable transition, and using real-time generated random numbers and corresponding trigger probabilities, determining whether each enableable transition in this simulation is... If a transition is not triggered, and for a transition that is determined to be triggered, update the token distribution in the input and output places of the transition to update the state of the Petri net dynamic model until there are no more enableable transitions; wherein, an enableable transition is a transition in which all input places have an initial token; the input place is the place at the source of the corresponding directed arc; the output place is the place at the end of the corresponding directed arc; determine the final token distribution as the final system state of this simulation, and determine the token transition path in this simulation as the state transition path of this simulation.

[0088] For a more detailed explanation of the working principle and procedures of this embodiment, please refer to the relevant description in Embodiment 1.

[0089] This invention, through a scenario database construction module 201, transforms unstructured, multi-source, heterogeneous raw data (history, equipment, operation, environment) into structured, computer-processable knowledge units with clear causal logic (state-cause-judgment). This provides the data and knowledge foundation for dynamic simulation and handling of new scenarios, enabling simulations to no longer rely on specific historical cases or the causal logic of single faults, but rather on general state transition logic. Through a dynamic model construction module 202, the scenario knowledge base is converted into a dynamic system model, accurately describing concurrent, asynchronous, and random behaviors, providing a precise computable model framework for subsequent large-scale, automated, and repeatable simulations. Through a power grid fault simulation module 203, each uncertain link (transition) in the model is given a quantified trigger probability, and all possible evolutionary branches are explored through random simulation. The probability distribution and specific evolutionary paths of various evolutionary final states are obtained through simulation, thereby improving the comprehensiveness of the simulation exploration of each cascading fault.

[0090] Example 3: This embodiment provides a terminal device, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus; The memory is used to store at least one executable instruction that causes the processor to perform the operation of the scenario-based power grid fault simulation method as described above.

[0091] Example 4: This invention provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device or apparatus containing the computer-readable storage medium to perform the scenario-evolution-based power grid fault simulation method as described above.

[0092] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0093] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A power grid fault prediction method based on scenario evolution, characterized in that, include: Based on the pre-acquired and stored multi-source heterogeneous data of the target power grid, a risk evolution scenario library for the target power grid is constructed. The multi-source heterogeneous data includes historical accident data, equipment parameter data, power grid operation data, and external environment data. The risk evolution scenario library includes several scenario triplets. Each scenario triplet consists of a state, a cause, and a judgment. The state represents the operating state or fault condition of the target power grid; the cause is the factor that causes the corresponding state to change; and the judgment is the conditional branch that causes the corresponding state to change. The risk evolution scenario library is mapped to a corresponding Petri net dynamic model; wherein, the Petri net dynamic model includes several locations, transitions, and directed arcs; the locations are obtained by mapping the states in the scenario triples; the transitions are obtained by mapping the judgments in the scenario triples and setting trigger conditions or attribute weights according to the causes in the scenario triples; the directed arcs are used to connect the locations and transitions according to preset logical associations; For each transition in the Petri net dynamic model, a trigger probability is calculated and set, and several random cascading failure evolution simulations are performed on the Petri net dynamic model with the trigger probability to obtain the cascading failure inference results of the target power grid; wherein, the trigger probability is calculated based on physical laws or historical decision data.

2. The power grid fault prediction method based on scenario evolution as described in claim 1, characterized in that, The step of constructing a risk evolution scenario database for the target power grid based on the pre-acquired multi-source heterogeneous data of the target power grid specifically involves: Natural language processing technology is used to extract key accident data from the historical accident data to form a structured historical event record; wherein, the key accident data includes power grid failure events, failure causes and maintenance operations; Based on the preset power grid operation procedures, and using the equipment parameter data and power grid operation data, power grid operation logic analysis is performed to obtain the operation rules of the target power grid. By using a preset association rule mining algorithm, the temporal correlation between the power grid fault events is analyzed based on the equipment parameter data, power grid operation data, and external environment data, so as to obtain the event association rules of the target power grid. Based on the historical event records and operating rules, several scenario triplets are constructed, and each scenario triplet is logically connected according to the event association rules to construct the risk evolution scenario library of the target power grid.

3. The power grid fault prediction method based on scenario evolution as described in claim 2, characterized in that, The aforementioned risk evolution scenario library is mapped to the corresponding Petri net dynamic model, specifically as follows: Map the state in each of the scenario triples to a place; The judgments in each scenario triple are mapped to one or more transitions, and trigger conditions or attribute weights are set for the transitions based on the reasons in the corresponding scenario triples. Based on the event association rules, the logical associations between the scenario triples in the risk evolution scenario library are determined, and based on the logical associations, the libraries and transitions are connected by directed arcs to construct an initial Petri net dynamic model. The initial Petri net dynamic model is subjected to reachability verification to obtain the final Petri net dynamic model.

4. The power grid fault prediction method based on scenario evolution as described in claim 1, characterized in that, Calculate and set the trigger probability for each transition in the Petri net dynamic model, including: Using a pre-acquired set of objective parameter sources, each transition in the Petri net dynamic model is sequentially matched, and the matched objective data source is determined as the trigger probability of the corresponding transition. The set of objective parameter sources includes the success rate of historical protection actions, the probability of equipment failure, and the probability of coupling risks. The matching operation includes keyword matching and type mapping.

5. The power grid fault prediction method based on scenario evolution as described in claim 4, characterized in that, After determining the objective data source for the matching as the corresponding trigger probability of the transition, the following is included: For transitions that are not matched with objective data sources, a cloud model is generated and associated with each transition based on pre-acquired historical decision data using a preset knowledge modeling algorithm; wherein, the cloud model includes the expected trigger probability, trigger probability entropy, and trigger probability hyper-entropy of the transition.

6. The power grid fault prediction method based on scenario evolution as described in claim 1, characterized in that, For the Petri net dynamic model with triggering probabilities, several random cascading failure evolution simulations are performed to obtain the cascading failure inference results of the target power grid, specifically as follows: Using the Monte Carlo simulation method, several random cascading failure evolution simulations were performed on the Petri net dynamic model with triggering probabilities, and the state transition path and corresponding final system state of each random cascading failure evolution simulation were obtained to obtain the failure evolution simulation log. Based on the fault evolution simulation log, risk analysis is performed on each state transition path, the final system state, and the transitions in the state transition path to obtain the risk analysis results. By integrating the fault evolution simulation logs and the corresponding risk analysis results, the cascading fault projection results of the target power grid are obtained.

7. The power grid fault prediction method based on scenario evolution as described in claim 6, characterized in that, Using the Monte Carlo simulation method, several simulations of stochastic cascading failure evolution were performed on the Petri net dynamic model with trigger probabilities. The process of performing a single cascading failure evolution simulation is as follows: The Petri net dynamic model is initialized based on the pre-stored initial fault setting data to determine the initial token distribution for each location. The algorithm iterates through each enableable transition, using real-time generated random numbers and their corresponding trigger probabilities to determine whether each enableable transition is triggered in the current simulation. For transitions that are determined to be triggered, the token distribution in the input and output places of the transition is updated to update the state of the Petri net dynamic model, until no enableable transitions remain. Here, an enableable transition refers to a transition in which all input places have an initial token; an input place refers to a place located at the source of the corresponding directed arc; and an output place refers to a place located at the end of the corresponding directed arc. The final token distribution is determined as the final system state of this simulation, and the token transfer path in this simulation is determined as the state transition path of this simulation.

8. A power grid fault simulation device based on scenario evolution, characterized in that, It includes a scenario library construction module, a dynamic model construction module, and a power grid fault simulation module, among which, The scenario library construction module is used to construct a risk evolution scenario library for the target power grid based on pre-acquired multi-source heterogeneous data of the target power grid. The multi-source heterogeneous data includes historical accident data, equipment parameter data, power grid operation data, and external environment data. The risk evolution scenario library includes several scenario triplets. Each scenario triplet consists of a state, a cause, and a judgment. The state represents the operating state or fault condition of the target power grid; the cause is the factor that causes the corresponding state to change; and the judgment is the conditional branch that causes the corresponding state to change. A dynamic model construction module is used to map the risk evolution scenario library to a corresponding Petri net dynamic model; wherein, the Petri net dynamic model includes several locations, transitions, and directed arcs; the locations are obtained by mapping the states in the scenario triples; the transitions are obtained by mapping the judgments in the scenario triples and setting trigger conditions or attribute weights according to the causes in the scenario triples; the directed arcs are used to connect the locations and transitions according to preset logical associations; The power grid fault simulation module is used to calculate and set the trigger probability for each transition in the Petri net dynamic model, and to perform several random cascading fault evolution simulations on the Petri net dynamic model with the trigger probability to obtain the cascading fault simulation results of the target power grid; wherein the trigger probability is calculated based on physical laws or historical decision data.

9. A terminal device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation of the scenario evolution-based power grid fault simulation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device or apparatus containing the computer-readable storage medium to perform the power grid fault simulation method based on scenario evolution as described in any one of claims 1 to 7.

Images