C3 ENERGY MACHINE LEARNING TEAM LEADER IN GEFCOM2014
C3 ENERGY MACHINE LEARNING TEAM LEADER IN GEFCOM2014 Jan 2015
Company data scientists place among top teams across categories
Jan 27, 2015 - C3 Energy’s preeminent machine learning team is the only team to earn a leading position across three categories in the provisional leaderboard for the Global Energy Forecasting Competition (GEFCom2014) after 12 weeks of competition. GEFCom2014 is an international forecasting competition sponsored by IEEE Power & Energy Society.
More than 40 teams from across the globe competed in GEFCom2014 in four tracks covering electric load, electricity price, wind power, and solar power forecasting. Each week the competition organizers released incremental data for the teams to aggregate, process, and create forecasts for the next period. GEFCom2014 is the first forecasting competition in the power and energy industry that asks for probabilistic forecasts. For each step throughout the forecasting horizon, the contestants were required to submit 99 forecasted quantiles instead of a single point estimate.
C3 Green Team, which consisted of six of C3 Energy’s data scientists, was one of the only teams to compete across all four categories and was the only team to place among the top five in three categories—electricity price, wind power, and solar power forecasting.
“C3 Energy has developed some of the most sophisticated applications of machine learning and forecasting techniques for today’s modern utility systems,” said S. Shankar Sastry, Dean of Engineering, UC Berkeley. “For GEFCom2014, the machine learning team leveraged this experience to create leading forecasts across critical areas to the energy ecosystem.”
The final leaderboard will be announced at the IEEE Power and Energy Society General Meeting in Denver, Colo., from July 26 – 30, 2015, where the top five teams will present their methodologies and the top three teams in each category will be honored.