Academic

Seminars Events Calendar


Filter by
Jump to
Search

Generating Long-Term Wind Scenarios Condtioned on Sequential Short-Term Forecasting Using LTM

Mr Kyle PerlineSchool of Civil & Environmental Engineering, Cornell University

Date:14 August 2017, Monday

Location:S16-06-118, DSAP Seminar Room

Time:04:00pm - 05:00pm

In this paper a new Long Term Generation (LTG) method is developed that generates long-term synthetic
wind scenarios conditioned on sequential short-term forecasts, and the Joint Distribution Comparison (JDC) method is developed to evaluate these scenarios. Scenarios generated using the LTG method can help to better evaluate the
operation of power systems with wind integration by providing additional forecast and outcome scenarios on which the systems can be backtested. The LTG method is distinct from the existing wind scenario generation methods, which
either generate short-term scenarios conditioned on a single short-term forecast or generate long-term scenarios that are not conditioned on any forecast information. The LTG algorithm is a generalization of the short-term scenario generation methods, where the marginal distributions of wind power at each time step conditioned on the forecasts are first estimated and then the full joint distribution is constructed. The JDC is a statistical test that compares the historical forecast and synthetic scenario joint distribution to the historical forecast and historical outcome joint distribution. These two new methods are applied to a dataset of historical wind forecasts and outcomes from Bonneville Power Administration.