CN-122021086-A - Scene evolution-based power grid fault deduction method, device, equipment and medium

CN122021086ACN 122021086 ACN122021086 ACN 122021086ACN-122021086-A

Abstract

The invention discloses a power grid fault deduction method, device, equipment and medium based on scenario evolution, wherein the method comprises the steps of constructing a risk evolution scenario library of a target power grid according to multisource heterogeneous data of the target power grid obtained in advance; the multi-source heterogeneous data comprises historical accident data, equipment parameter data, power grid operation data and external environment data, the risk evolution scenario library comprises a plurality of scenario triplets, the scenario triplets comprise states, reasons and judgments, the risk evolution scenario is mapped into a corresponding Petri network dynamic model, trigger probability is calculated and set for each transition in the Petri network dynamic model, and the Petri network dynamic model is subjected to random cascading failure evolution simulation for a plurality of times to obtain cascading failure deduction results of a target power grid. The invention can dynamically deduce cascading faults even in a brand new risk scene caused by lack of historical data support or multi-factor complex coupling, and improves the generalization capability of the power grid cascading fault deduction.

Inventors

YANG YIMING
TAN JING
Ji Zheyi
MI YI
WANG CHEN
LI ZHEQI
WANG LIN
WANG YAPING
YU JUANYING
LI HONGJIANG
CHEN YINGHUA
XU XUESONG
GE YANQIN
JIANG YINGHAN
ZHANG TIANMI
CHE MENG
LI XUEFENG
KONG JUAN
FENG XIN
Liu Yaomei
WANG HAIYOU
ZHANG YANG
Cui xue
ZHU WEI
LU QIANG
Yi Xiying

Assignees

国网经济技术研究院有限公司
国网能源研究院有限公司

Dates

Publication Date: 20260512
Application Date: 20260119

Claims (10)

1. The utility model provides a power grid fault deduction method based on scene evolution, which is characterized by comprising the following steps: The risk evolution scene library of the target power grid is constructed according to the pre-acquired and stored 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 triples consist of states, reasons and judgment, the states are operation states or fault conditions of the target power grid, the reasons are factors causing corresponding state changes, and the judgment is a conditional branch causing corresponding state changes; mapping the risk evolution scene library into a corresponding Petri network dynamic model, wherein the Petri network dynamic model comprises a plurality of libraries, transitions and directed arcs, wherein the libraries are obtained by mapping states in the scene triplets, the transitions are obtained by judging and mapping in the scene triplets and setting triggering conditions or attribute weights according to reasons in the scene triplets, and the directed arcs are used for connecting the libraries and the transitions according to preset logic association; Calculating and setting a trigger probability for each transition in the Petri network dynamic model, and carrying out random cascading failure evolution simulation for a plurality of times on the Petri network dynamic model with the trigger probability to obtain a cascading failure deduction result of the target power grid, wherein the trigger probability is calculated according to a physical rule or historical decision data.
2. The method for deducing a power grid fault based on scenario evolution according to claim 1, wherein the constructing a risk evolution scenario library of a target power grid according to the pre-acquired multi-source heterogeneous data of the target power grid specifically comprises: Extracting accident key data from the historical accident data through a natural language processing technology to form a structured historical event record, wherein the accident key data comprises a power grid fault event, a fault reason and maintenance operation; according to a preset power grid operation rule, carrying out power grid operation logic analysis based on the equipment parameter data and the power grid operation data to obtain an operation rule of the target power grid; Analyzing time sequence association among the grid fault events according to the equipment parameter data, the grid operation data and the external environment data by a preset association rule mining algorithm to obtain event association rules of the target grid; And constructing a plurality of scenario triples according to the historical event records and the operation rules, and logically connecting each scenario triplet according to the event association rules to construct a risk evolution scenario library of the target power grid.
3. The method for power grid fault deduction based on scenario evolution according to claim 2, wherein the risk evolution scenario library is mapped into a corresponding Petri network dynamic model, specifically: mapping the state in each scene triplet to a library; Mapping the judgment in each scene triplet to one or more transitions, and setting triggering conditions or attribute weights for the transitions according to reasons in the corresponding scene triples; determining logic association among all scene triples in the risk evolution scene library according to the event association rule, and connecting all libraries and transitions through directed arcs according to the logic association to construct an initial Petri network dynamic model; And carrying out reachability verification on the initial Petri network dynamic model to obtain a final Petri network dynamic model.
4. The scenario evolution-based power grid fault deduction method according to claim 1, wherein calculating and setting a trigger probability for each transition in the Petri network dynamic model comprises: And sequentially carrying out matching operation on each transition in the Petri network dynamic model through a pre-acquired objective parameter source set, and determining that the matched objective data source is the triggering probability of the corresponding transition, wherein the objective parameter source set comprises a historical protection action success rate, an equipment failure probability and a coupling risk probability, and the matching operation comprises keyword matching and type mapping.
5. The scenario evolution-based power grid fault deduction method according to claim 4, wherein after determining that the objective data source on the match is the trigger probability of the corresponding transition, comprising: And generating and associating respective cloud models for the transitions of the unmatched objective data source according to the pre-acquired historical decision data through a preset knowledge modeling algorithm, wherein the cloud models comprise trigger probability expectations, trigger probability entropy and trigger probability super entropy of the corresponding transitions.
6. The method for power grid fault deduction based on scenario evolution according to claim 1, wherein the method for power grid fault deduction based on scenario evolution is characterized in that the dynamic model of the Petri network with trigger probability is subjected to random cascading fault evolution simulation for several times to obtain cascading fault deduction results of the target power grid, specifically: Carrying out a plurality of random cascading failure evolution simulations on the Petri network dynamic model with the trigger probability by a Monte Carlo simulation method, and obtaining a state transition path and a corresponding final system state of each random cascading failure evolution simulation to obtain a failure evolution simulation log; according to the fault evolution simulation log, risk analysis is respectively carried out on each state transition path, the final system state and transition in the state transition path, and a risk analysis result is obtained; and integrating the fault evolution simulation log and the corresponding risk analysis result to obtain a cascading fault deduction result of the target power grid.
7. The power grid fault deduction method based on scene evolution according to claim 6, wherein the Petri network dynamic model with trigger probability is subjected to several times of random cascade fault evolution simulation by a Monte Carlo simulation method, wherein the process of single cascade fault evolution simulation is specifically as follows: initializing the Petri network dynamic model according to pre-stored initial fault setting data, and determining initial token distribution of each library; Traversing each possible transition, judging whether each possible transition in the simulation is triggered or not through a random number generated in real time and a corresponding trigger probability, and updating token distribution in an input warehouse and an output warehouse of the transition for determining the triggered transition so as to update the state of the Petri network dynamic model until no possible transition exists, wherein the possible transition refers to the transition that all input warehouses are provided with initial tokens; and determining the final token distribution as the final system state of the current simulation, and determining the token transfer path in the current simulation as the state transfer path of the current simulation.
8. The utility model relates to a power grid fault deduction device based on scene evolution, which is characterized by comprising a scene library construction module, a dynamic model construction module and a power grid fault deduction module, wherein, The system comprises a scenario library construction module, a risk evolution scenario library, a condition analysis module and a condition analysis module, wherein the scenario library construction module is used for constructing a risk evolution scenario library of a target power grid according to pre-acquired 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 scenario library comprises a plurality of scenario triples, the scenario triples consist of states, reasons and judgment, the states are operation states or fault conditions of the target power grid, the reasons are factors which cause the corresponding states to change, and the judgment is a conditional branch which causes the corresponding states to change; The dynamic model construction module is used for mapping the risk evolution scene library into a corresponding Petri network dynamic model, wherein the Petri network dynamic model comprises a plurality of libraries, transitions and directed arcs, the libraries are obtained by mapping states in the scene triad, the transitions are obtained by judging and mapping in the scene triad and setting triggering conditions or attribute weights according to reasons in the scene triad, and the directed arcs are used for connecting the libraries and the transitions according to preset logic association; The power grid fault deduction module is used for calculating and setting triggering probability for each transition in the Petri network dynamic model, and carrying out random cascading fault evolution simulation for a plurality of times on the Petri network dynamic model with the triggering probability to obtain a cascading fault deduction result of the target power grid, wherein the triggering probability is calculated according to a physical rule or historical decision data.
9. The terminal equipment is characterized by comprising a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus; The memory is configured to store at least one executable instruction that causes the processor to perform the operations of the scenario evolution-based power grid fault deduction method according to any one of claims 1 to 7.
10. A computer readable storage medium, characterized in that the computer readable storage medium comprises a stored computer program, wherein the computer program, when run, controls a device or apparatus in which the computer readable storage medium is located to perform the scenario evolution based power grid fault deduction method according to any one of claims 1 to 7.

Description

Scene evolution-based power grid fault deduction method, device, equipment and medium Technical Field The invention relates to the technical field of power grid operation and maintenance management, in particular to a power grid fault deduction method, device, equipment and medium based on scene evolution. Background With the large-scale grid connection of new energy sources such as wind power, photovoltaic and the like and the wide application of direct current transmission and power electronic equipment in a power grid, the structure and the operation characteristics of a power system are increasingly complex, and the risk of cascading faults is obviously increased. The cascading failure is usually triggered by initial disturbance, and propagates step by step through multiple mechanisms such as electric coupling, protection action, information interaction and the like, and can cause serious consequences such as large-area power failure and the like. In order to improve the security defense capability of the power grid, early warning and accurate prevention and control of faults are realized, and it is important to develop power grid fault deduction. The fault deduction aims at simulating the whole process from occurrence and development to final results of the fault, and reveals an evolution path and key links of the fault, so that decision basis is provided for operators, emergency plans are optimized, and system operation risks are reduced. At present, a plurality of researches exist in the field of power grid cascading failure deduction and risk assessment, and the existing method mainly relies on historical operation data, failure records or simulation cases, and constructs a failure propagation model and performs risk analysis through technologies such as statistical learning, causal deduction or attack graph modeling. For example, some methods extract causal relationships between an initial fault and a subsequent fault by using machine learning models such as logistic regression and neural networks based on a large number of historical cascading failure scenes, combine time sequence features and sample to generate a failure scene, or model propagation paths of all the faults, and finally perform risk assessment. However, such methods rely essentially on historical data or statistical laws underlying known attack patterns to be causally related, with deductive capabilities limited to the scope of historical experience. When the power grid faces novel risk scenes beyond the history record, namely extreme weather combination, novel equipment fault mode or linkage effect brought by multi-factor complex coupling which are not seen in the past, the existing method is difficult to carry out effective and reliable dynamic deduction and root cause tracing on the brand new scenes due to lack of corresponding historical data support, so that early warning and decision support capability of the power grid in coping with unknown and uncertain risks is limited. Disclosure of Invention The invention provides a method, a device, equipment and a medium for power grid fault deduction based on scenario evolution, which can dynamically deduct cascading faults even though a brand new risk scene is lack of historical data support or is caused by multi-factor complex coupling through the scenario evolution and dynamic simulation, and improve the generalization capability of the power grid cascading fault deduction. In a first aspect, an embodiment of the present invention provides a scenario evolution-based power grid fault deduction method, including: The risk evolution scene library of the target power grid is constructed according to the pre-acquired 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 triples consist of states, reasons and judgment, the states are operation states or fault conditions of the target power grid, the reasons are factors causing corresponding state changes, and the judgment is a conditional branch causing corresponding state changes; mapping the risk evolution scene library into a corresponding Petri network dynamic model, wherein the Petri network dynamic model comprises a plurality of libraries, transitions and directed arcs, wherein the libraries are obtained by mapping states in the scene triplets, the transitions are obtained by judging and mapping in the scene triplets and setting triggering conditions or attribute weights according to reasons in the scene triplets, and the directed arcs are used for connecting the libraries and the transitions according to preset logic association; Calculating and setting a trigger probability for each transition in the Petri network dynamic model, and carrying out random cascading failure evolution sim