Collaborative Research: An Intelligent Restoration System for a Self-healing Smart Grid (IRS-SG)

Clemson University, Clemson SC

Investigators

Abstract

How can we restore power more effectively after a major outage, such as a hurricane, a cascade blackout or a more local outage? Can we use modern computer-based methods to get better performance than what we have today, in a system which is mainly informal and based on guesswork under conditions of great stress and limited information? This new collaborative project will bring together an expert on power restoration with a pioneer of new intelligent computation methods for the power grid, in hopes of finding a more powerful and modern way to restore power more effectively. The goals are; (1) to develop an online adaptive restoration tool using advanced scalable computational intelligence techniques; (2) investigate a novel scheme to use renewable resources in system restoration; (3) explore a blackstart unit investment strategy to improve the self-healing capability; and (4) real-time implementation and demonstration of the new system on benchmark and utility power systems. The grant will include travel to New Zealand, to discuss use of the new system to help provide better response to events like earthquakes. It will also include outreach to Native Americans in the Dakotas. The problem of efficient restoration is very difficult from a technical point of view. A small number of researchers, like the lead PI, have developed a few tools of practical use in this problem, but it is still largely an unsolved problem. The main justification for NSF involvement in this area, and for significant hope of success, is the use of intelligent systems methods far beyond what anyone has applied in the past to this problem, methods pioneered in the intelligent systems part of the EPCN program at NSF (described for example in the book Handbook of RLADP edited by Frank Lewis and Derong Liu). Data from new sources such as synchrophasors will be part of this work. The underlying algorithms are designed from the start to run on massively parallel distributed systems, such as Cellular Neural Network hardware, which allow much faster real-time computation than traditional sequential computers and algorithms.