We’re nearing the halfway mark on the AMS 2013-2014 Solar Energy Prediction Contest. Right now the leaderboard is headed largely by teams that participate the most, but there’s still time to join in if you want to have a chance at a prize.
The 130-day contest opened on July 8 and closes on November 15. It is run by the AMS Committees on Artificial Intelligence Applications to Environmental Science, Probability and Statistics, and Earth and Energy, which began reaching out to a wider community of participants last year by hosting the contest on the website, Kaggle. More than 90 teams are trying their skill with statistical and machine learning techniques. The daily challenge is to predict incoming solar energy at a preselected set of 98 Oklahoma Mesonet sites that serve as a virtual solar farm.
Hopefully, the contest will show “which statistical and machine learning techniques provide the best short-term predictions of solar energy production,” say the organizers, Amy McGovern and David John Gagne II of the University of Oklahoma. “Power forecasts typically are derived from numerical weather prediction models, but statistical and machine learning techniques are increasingly being used in conjunction with the numerical models to produce more accurate forecasts.”
The contest website provides input numerical weather prediction data from the NOAA/ESRL Global Ensemble Forecast System (GEFS) Reforecast Version 2. Data include all 11 ensemble members and the forecast time steps 12, 15, 18, 21, and 24 provided in netCD4 files. There’s also a training dataset and a public testing dataset. Basically all you have to do is jump right in and see how well your methods work.
Teams identified as preliminary winners (based on mean absolute error, a common metric in the renewable energy industry) will submit their code to the judges after the contest for verification. The verified winners will make their code open-source, and will be honored (and expected to present) at the AMS Annual Meeting in February 2014 in Atlanta.
The prizes are sponsored by EarthRisk Technologies, Inc., and include $500 awarded to the winner, $300 for second place, and $200 for third. Each of the top three gets their abstract fees waived for the upcoming AMS Annual Meeting, and the top student forecaster gets both abstract fee and conference registration waived.
Daily solar energy data were provided by the Oklahoma Mesonet with the assistance of Jeffrey Basara of the Oklahoma Climatological Survey. The GEFS Reforecast Version 2 data were developed and provided by Thomas Hamill of NOAA.