Reliable operation of cyber-physical systems (CPS) of societal importance such as Smart Electric Grids is critical for the seamless functioning of a vibrant economy. Sustained power outages can lead to major disruptions over large areas costing millions of dollars. Efficient computational techniques and tools that curtail such systematic failures by performing fault diagnosis and prognostics are therefore necessary. In this project, we have developed modeling formalisms, algorithms, and prototype tools for: 1. Timely diagnosis of the failures of physical and cyber components in the power system. 2. Identifying the potential mis-operations of protection systems and automatic controls using available information from the physical and the cyber components of this system. 3. Providing prognosis of the system at any time, given its current state after a failure.

The intellectual merit of the proposed approach is in that it does not involve complex real-time computations on high-fidelity models, but performs reasoning using efficient graph algorithms based on the observation of various anomalies in the system. Therefore, this approach can uniquely account for failures, not only in the physical but also in the cyber subsystems that tightly interact in this CPS.

The broader impact of this work are the new models, algorithms, and tools to find the root causes of failures as well as failure prognosis in heterogeneous electrical CPS. The outcome in terms of better diagnostic/ prognostic tools from the proposed research will help in developing better failure analysis approach, which in future can help prevent serious fault cascades such as blackouts. In the life of the project, the team has directly involved a number of graduate students in the research effort, who were integral to delivering the listed research outcomes. In the context of this work, the students interacted with their counterparts in computer science, presenting a unique training environment for power systems engineers.

Collectively, the team collaborated to propose the diagnosis of fault events and prognosis of cascade failures. The models were developed and validated in a shorter cycle, and produced much more details in terms of relay responses, which can be used to improve the analysis. Specifically, NCSU team:

1. Developed dynamic models for Generator machines, and respective excitation and governors which allow stable steady state operation and model cascaded events well. These models allowed more details on system level analysis of cascaded faults.

2. Implemented models of commercial distance relays in the system. This allowed a detailed insight into the cascade behaviour and possible reasons for progression.

3. Implemented full scale distance relay based protection for IEEE 14 bus system to capture system wide dynamics given an event. This was not possible before. We were able to model progressive faults in multiple cascade stages (Slow and Fast)

4. Identified the role of a bus fault relay misoperation in escalating a cascade. A demo was developed for a WSCC 9-bus system where a bus fault relay misoperation converts a regular line fault into a cascaded blackout. We developed a demonstration case based on WSCC 9-bus system. Here, a bus relay misoperated due to a cyber fault on a local line relay causing escalation of a line fault into a cascaded blackout. This was identified as a potential venue for future research.

5. Reduced the development and validation time by more than half, which improving on the level/fidelity of results.

6. Validated the results of Cascaded failure events with and without fault mitigation strategies.

Last Modified: 02/23/2018
Modified by: Srdjan M Lukic