You are here

ISE Welcomes Dr. Bita Analui on 3/9/17

Overcoming model uncertainty and inaccuracies in power system planning and operations

Seminar by Bita Analui, Ph.D., Postdoctoral Scholar

School of Electrical, Computer, and Energy Engineering; Arizona State University

Thursday, March 9th from 5:30 – 6:30 pm

210E Baker Systems, 1971 Neil Avenue

Planning the operations of power system introduces a number of optimization problems that need to be solved in different time resolutions under various sources of uncertainty. Due to the inherent non-linearities and non-convexities, these problems are often very challenging to solve at large scales. Multistage stochastic optimization is the state of the art tool to address sequential decision-making under uncertainty. This talk focuses on addressing and overcoming inaccuracies that are tied to the practical application of this tool.

First, uncertainty is not only due to a probabilistic model but also data-driven, as there is an intrinsic ambiguity in the choice of a probability model that may result in misleading solutions. To address this problem, we first present a theoretical framework to address model uncertainty (ambiguity) in multistage stochastic optimization problems and the corresponding algorithms to compute the saddle point solution.  We present an application of this method to a multistage stochastic economic dispatch incorporating ramping constraint, in which reserve requirements can be represented as distributionally robust scenarios of generation schedules.  In the second part of this talk, we address the issue of selecting scenario trees and solving problems dynamically. We show how one can formulate dynamic multistage stochastic unit commitment in which scenarios are selected from a library of scenario trees that is also dynamically updated as new data are observed. Finally, a continuous-time formulation for multistage stochastic unit commitment is proposed to capture the continuous nature of net-load and optimally schedule continuous-time generation and ramping scenarios, further reducing the inaccuracy that is associated with solving a discrete problem rather that the true variational problem.

Bita Analui is a postdoctoral scholar in Electrical, Computer and Energy Engineering School and Postdoc in Computing Best Practice Fellow at Arizona State University. Prior to joining ASU, Bita was a research assistant at Computational Optimization Doctoral College, University of Vienna in Austria, where she received her Ph.D. in Statistics and Operations Research and graduated cum laude in 2014. Her thesis contribution was on the theory and solution algorithms for multistage stochastic optimization problems under model uncertainty. Bita’s research interests include stochastic optimization, data analytics, statistical modelling, scenario generation and designing dynamic decision models under uncertainty for real world class of problems. Much of her recent works span all of these fields in order to provide a new perspective towards conventional problems.