248 Benchmark Solar Power Forecasts

Monday, 7 January 2019
Hall 4 (Phoenix Convention Center - West and North Buildings)
William F. Holmgren, The Univ. of Arizona, Tucson, AZ; and A. T. Lorenzo, C. W. Hansen, J. Sharp, and A. Tuohy

The DOE SETO Solar Forecasting 2 program and the broader solar power forecasting community seeks to measure the skill of new forecasts against benchmark forecasts to better understand the relative merits of forecast techniques. The benchmark forecasts can be created using a variety of methods, each with their own tradeoffs. Furthermore, each benchmark forecast may illuminate or hide particular positive or negative aspects of the new forecast. It is therefore valuable to examine candidate benchmark solar power forecast methods in detail. This work explores the attributes, techniques, and timescales associated with candidate benchmark forecast methods. Desirable attributes of a benchmark forecast include low cost, data and forecast availability, and transparency. We will examine forecasts derived from numerical weather models, forecasts derived from statistical methods, and combinations of these for forecast horizons of minutes to one week. This work is supported by the DOE SETO Solar Forecasting 2 program.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner