University of Minnesota
University Relations
http://www.umn.edu/urelate
612-624-6868
myU OneStop


Go to unit's home.

Home | Seminars and Symposia | Past seminars/symposia: Wednesday, March 7, 2012

DTC Leading Edge Seminar Series

Machine Learning Problems in Renewable Energy

by

Marc Light, Sriharsha Veeramachaneni, Ken Williams
WindLogics, Inc.

Wednesday, March 7, 2012
3:30 p.m. reception
4:00 p.m. seminar

401/402 Walter Library

Marc Light

Marc Light

Marc Light

Sriharsha Veeramachaneni

Solar and wind energy are highly variable resources, in contrast to relatively constant generation from fossil fuels. In order to maintain the reliability of the electrical grid, operators require forecasts of future energy production. Power traders also use such forecasts to help achieve favorable market trading conditions. At the other end of the wire, the electrical load must also be forecast because of the requirement to balance generation and load at all times. These forecasts depend on numerical weather prediction and other inputs. These and related forecasting tasks present interesting machine learning challenges such as the modeling of noise in the predictive variables (often weather variables themselves predicted), functional regression as opposed to conventional multivariate regression, and hierarchical modeling to handle turbine, wind farm, and grid-level fleet forecasts. Often physically-motivated modeling can only take us part of the way towards forecasting the target variable, necessitating machine learning and statistical techniques to bridge the final gap.

 

Marc Light joined WindLogics in June of 2011 as a senior research scientist. Previously he held research positions at Thomson Reuters, University of Iowa, MITRE, University of Stuttgart, and University of Tuebingen. At the University of Rochester, he completed his PhD in the area of natural language processing. His research interests include stochastic sequence labeling and tools for feature engineering and selection. Sriharsha Veeramachaneni received his Ph.D. in computer engineering from the Rensselaer Polytechnic Institute, New York in 2002. He is currently a senior research scientist at WindLogics, Inc. His research interests include Bayesian statistics, machine learning, and information theory.

WilliamsKen Williams is a Senior Research Scientist in the Applied Math group at WindLogics, Inc., specializing in Machine Learning, Data Mining, and Reproducible Research. He earned his Master's degree from the University of Sydney in Computer Science and Natural Language Processing. He has worked in the Research and Development group at Thomson Reuters and as an independent consultant.