DIMACS Workshop on Algorithmic Decision Theory for the Smart Grid

October 25 - 27, 2010
DIMACS Center, CoRE Building, Rutgers University

K. R. Krishnan, Telcordia, krk at research.telcordia.com
Linda Ness, Telcordia, lness at telcordia.com
Tom Reddington, Bell Labs, treddington at research.bell-labs.com
Presented under the auspices of the Special Focus on Algorithmic Decision Theory.

Stable, secure and reliable operation of the national electric power grid is critical to national security and the nation's economy. Yet, today's electric power systems are vulnerable to major disruptions, as has been shown by recent black-outs, with substantial economic impact and human suffering. Decisions about the design, operation, and repair of these systems must be made rapidly using massive amounts of data that new methods and tools allow us to gather about the state of the system. This Workshop will investigate the Algorithmic Decision Theory (ADT) challenges in the management and control of the electric power grid. These include the scale of the problem, the intrinsic uncertainties in monitoring its ``state'', and the potential for triggering cascading blackouts. We will examine how the grid can benefit from the use of formal methods of ADT and how such methods need to be adapted and enhanced to deal with the size and complexity of the grid as it grows to accommodate millions of Intelligent Meters, Electric Vehicles, and diverse classes of energy sources.

In October of 2009, President Obama announced $3.4B of funding for modernization of the US power grid. A report by the National Academies of Science issued in 2009, prepared as part of their America's Energy Future study, concluded that grid modernization is essential to increasing US usage of renewable energy. This increased attention highlights the timeliness of the proposed Workshop, which will bring together researchers and experts in ADT, complex adaptive systems, security and Smart Grid operation to identify research challenges in ADT whose solutions will be crucial to the successful realization of the Smart Grid's goals of an ultra-efficient, self-healing, attack-resistant grid, capable of utilizing a wide range of generation and storage options to reliably meet 21st century power needs.

Owing to the intrinsic uncertainty in fluctuations in demand for power and the inherent possibility of measurement errors, the "state'' information that is collected is fraught with a degree of unavoidable uncertainty. In this environment, there is need to estimate uncertainty and consider the risk of inaction or wrong action due to unreliable data (aggravated by the fact that some controls might be "safe'' in one operating regime and unsuitable in another). Since control decisions are based on "cognition'' of many streams of data, real-time algorithms operating on multiple streams of noisy data are required. In addition, some decisions need to be made collectively and in a distributed fashion, which adds to the complexity. Thus, algorithms and decision making are involved in all these steps.

Questions to be addressed in the Workshop include:

(a) How do the algorithmic challenges for decision theory for the power grid (involving situational awareness, stability, resilience to cyber attacks, etc.) differ from those for telecommunication and computer networks, and what factors contribute to those differences? (b) What are the challenges in fusing multiple (possibly incomplete and/or inconsistent) data streams from monitors into a reduced consistent set suitable for action and for human interpretation? (c) What is the time-scale in which diagnostic decisions must be reached in order for effective action against the onset of instability? (d) What are the indicators of incipient instability in the grid and when is it advisable to "break'' the grid and isolate unstable segments in order to avoid cascading failures? (e) How vulnerable is the grid to disruption by the introduction of malicious software, and what countermeasures are available for detecting and defeating such intrusions? (f) What types of algorithms are needed to integrate alternative power sources and "smart'' appliances (which can be programmed to defer their power demands to non-peak hours) into the operation of the grid? (g) How can one diagnose instabilities in the power grid arising, not from equipment failure, but from aberrant behavior caused in the communications network, for example, by cyber attacks? (h) What privacy issues need to be addressed in the Smart Grid?

The electric power grid has lagged behind telecommunication and computer networks in the use of formal mathematical tools for its management, as noted in the Smart Grid literature. The possibilities for monitoring and digital control offer the opportunity to modernize grid operation by innovative control algorithms, keeping it operating in the presence of demand fluctuations and cyber-attacks. The Workshop will help expedite the development of algorithmic solutions to the problems of making real-time predictions and decisions for controlling and securing the grid as it grows in complexity with more distributed power sources and customers.

Next: Call for Participation
Workshop Index
DIMACS Homepage
Contacting the Center
Document last modified on April 14, 2010.