Failure Diagnosis about Cyber-Physical Energy Systems Using Temporal Causal Models

Abstract

In the light of recent attacks, making the power grid more resilient has become one of the top national priorities. One way of increasing the grid resiliency is by enriching the grid operators with efficient online management and decision support tools that improve their situational awareness to better handle the daily operations under different operating conditions. An important aspect of any online management system is its ability to quickly perform automated diagnosis to identify the root cause(s) of the aberrant system behavior. However, diagnosing faults and other events is challenging in a heterogeneous cyber physical system (CPS) such as the power grid that contains a large number of physical equipment, protection devices and actuators etc. Moreover, the interaction between physical equipment and protection devices along with latent defects in their embedded protection algorithms render the traditional approaches ineffective for modeling and diagnosing the effect of faults in cyber physical energy systems (CPES). To this end, we present a model-based diagnosis system aimed for safety critical CPS that uses a novel graphical modeling language, Temporal Causal Diagrams (TCD) to represent fault effect propagation while considering faults in power system apparatus, protection devices and the interaction between the two domains. The underlying diagnosis system uses a graph-theoretic search algorithm to reason about the state of the system in a swift manner. Specifically, in this work we make the following contributions 1) Developed the syntax and execution semantics of the new formalism. 2) Designed and validated TCD fault models for different components of the CPES 3) Designed a synthesis pipeline to generate the TCD fault model from a power network topology. 4) Developed and validated a hierarchical diagnosis engine to reason about observed faults effects while considering faulty and nominal protection system behaviors. 5) Designed and validated simple cascade prognostics and mitigation methodologies as applications supported by the proposed diagnosis system.