| In order to meet the construction requirements of the State Grid Corporation’s "big operation",the smart grid dispatching control system(D5000 for short)integrates more than 10 independent application systems inside a dispatching center into a basic platform and four types of application systems.(Real-time monitoring and early warning,dispatch planning,safety check and dispatch management).The system is based on a service-oriented architecture(Service Oriented Architecture,SOA)and uses multi-core computer cluster technology to organically combine a more decentralized and independent power grid dispatching system with an Internet service model to form a complex cyber-physical system.Once a component in the system fails,it will most likely affect its associated components,which will cause a series of alarms,which will bring great trouble to the operation and maintenance personnel’s fault location work.However,under the condition that the logical topology relationship is difficult to sort out,the existing method of fault tracing based on the logical topology diagram is no longer applicable.Using system operating data modeling to simulate system behavior and replacing "logical topology" with "functional topology" can provide new ideas for solving such problems.This paper studies the fault location method of cyber-physical system based on data-driven graph model.The main work is as follows:(1)The method of modeling undirected graphs of components in cyber-physical systems is studied,and a fault location method based on invariant network model and fault injection is proposed.First,use the system component operation data collected in the system log to build an invariant network model of the system,and then perform node or component fault injection and diffusion modeling to establish a fault network set.Second,according to the original time series Data conditions,formulate data quality evaluation rules to identify whether the data has changed suddenly;finally,use the actual fault network and fault network centralized fault network local fitting to perform fault tracing,and use the data quality evaluation rules to modify the results to achieve Pinpoint the source of system failures.(2)The directed graph modeling method and fault tracing method based on the new causality theory in the cyber-physical system are studied,and a series of fault location methods based on the dynamic hierarchical new causality network model are proposed.First,consider the potential one-way influence relationship between the time series in the information system,and use the new causality relationship to represent the dynamic relationship between the collected time series,which lays the foundation for simulating the one-way propagation of faults in the system.Second,comprehensive considering the requirements of different fault source tracing accuracy,the characteristics of time series collection location and system architecture were used to integrate the causal network model into a multi-layer structured network model.Finally,an algorithm based on the importance of the node is proposed to accelerate the strategy.Nodes with more non-zero elements in the reachability matrix are preferentially detected.Normal nodes with more reachability relationships are not processed in subsequent steps,thereby improving the efficiency of the fault source location algorithm.(3)The data-driven hybrid graph model modeling method for cyber-physical systems is studied,and a fault location method based on the hybrid graph model is proposed.First,establish a system’s invariant network model and a new causal network model separately;then,formulate the corresponding rules,build a system’s hybrid network model,and use the system’s hierarchical information to integrate the model into a clear hierarchical structured hybrid graph model;Finally,the characteristics of the fault link correlation graph model are analyzed to isolate the nodes that are highly correlated with the fault source. |
PDF Full Text Request