6th International Conference Energy & Meteorology: Abstract Submission

An investigation of dynamic selection of WRF PBL schemes for renewable energy forecasting in Ireland. (740)

Seanie Griffin 1 2 , Conor Sweeney 1 2 , Frank McDermott 1 3
  1. Energy Institute, University College Dublin, Dublin, Ireland
  2. School of Mathematics and Statistics, University College Dublin, Dublin, Ireland
  3. School of Earth Sciences, University College Dublin, Dublin, Ireland

Objective and Background 

Forecast models are used widely to predict renewable energy output for transmission system operators and energy market traders. Atmospheric turbulence has an impact on the forecasting of hub-height wind speeds and these processes are mostly represented using planetary boundary layer (PBL) schemes within forecast models. Six PBL schemes available in the Weather Research and Forecasting (WRF) model have been compared for renewable energy forecasting in Ireland.These forecasts have been run at 2.25km grid-spacing and driven by ECMWF IFS forecast data for each PBL scheme, as well as an adaptive forecasting system with a horizontal resolution of 750m. This study examines the factors which influence the performance of different PBL schemes and investigates whether this information can be used to produce a more accurate wind power forecast.

Twelve months of 2.25km forecasts have been compared to mast observations from a collection of wind-farms around Ireland and with surface observations from Irish synoptic weather stations. This study focuses on the 24-48 hour forecast horizon as this represents a relevant time period for day-ahead energy market trading. The overall forecast skill is examined to determine the most accurate PBL scheme for the forecasting of renewable energy output. The prediction of wind ramping events are also analysed.    Although a particular PBL scheme may have the best overall skill score, previous studies have found that the relative performance of different PBL schemes can vary depending on weather conditions, e.g. atmospheric stability. We investigate the use of a forecasting system with adaptive physics selection, where the PBL scheme for a single high-resolution (750m) forecast is dynamically selected based on the recent performance of the 2.25km forecasts. These adaptive forecasts are compared to 750m forecasts run with a static PBL scheme.    

Preliminary Results 

Preliminary analysis has shown that the best performing PBL does change over time for 10m wind speed at many synoptic weather stations and also for winds at height from wind-farm masts; with different stations often exhibiting a unique pattern of variability in PBL performance.