New Jersey Chapter

Date:
Thursday November 6, 2014
Agenda:
6:00-6:30 Mixer
6:30-7:30 Dinner
7:30-9:00 Presentation
Location:
100 Rockafeller Road, Piscataway, NJ 08854
Room 4095, Rutgers Business School, Livingston Campus, Rutgers University
Enter starting address to get directions: http://rumaps.rutgers.edu/location/lot-scarlet-lot-rac

| Home | INFORMS Subdivisions | INFORMS |

Chance-constrained generator dispatch

Daniel Bienstock
Columbia University

Daniel Bienstock is a professor at the Industrial Engineering and Operations Research, and Applied Math and Applied Physics departments at Columbia University, where he has been since 1989. His research focuses on high-performance methodologies and computing implementations of algorithms for optimization problems, with special focus on large nonconvex and integer programming models. A second focus is on applications of optimization to problems arising in electrical power transmission. Please visit http://www.columbia.edu/~dano/ for more information on Daniel Bienstock.

Abstract

A common process for managing electrical power transmission systems (aka "the grid") involves the process known as OPF, or Optimal Power Flow, which is run as often as every fifteen minutes (sometimes as often as every five minutes) so as to choose generator setpoints (power output) for the next time window, using demand estimates for that period, and with the goal of minimizing generation cost (a convex function of the generator outputs). Deviations of demand away from the predicted values are handled in real time using the so-called "secondary (frequency) response" framework, in which the underlying physics causes generators to marginally slow down or speed up in response to instantaneous load changes (we will explain this more fully in the talk). This overall setup has worked quite well over the last fifty years, with power produced efficiently and securely.

When incorporating renewable sources, especially wind, this setup does not work so well because real-time wind fluctuations can be quite large (in either direction) and the frequency response framework can cause line flows to become very large, a phenomenon already observed in regions with high renewable penetration (e.g. Germany). In this talk we describe a modification to the OPF computation that takes into account the intermittent nature of wind by formulating a chance-constrained (stochastic) optimization problem that can be solved very quickly. We will describe some of the underlying mathematics, the computational tricks needed to make it run fast, and provide some examples of the methodology in action.


Light dinner will be provided. Attendance at the lecture will be for NJ and NY Metro INFORMS members. 2014 chapter dues of $10 for INFORMS National Member, $11 for Non-member or $3.00 for student or retiree.

Further information can be obtained by contacting Keh-Wei Lih at lih@att.com. Visit NJ INFORMS Chapter home page at http://rutcor.rutgers.edu/~lih/informs/njchap.html

| Home | INFORMS Subdivisions | INFORMS |