Improve Power Generation Forecasting by Neural Networks & Time Series Forecasting

Improve Power Generation Forecasting by Neural Networks &
Time Series Forecasting

The Challenge

A wind farm in Europe, supplying power to the electric grid was having a high margin of error of 30% in its output prediction, resulting in high penalty to the grid. ( Errors greater than 10% incurred a penalty ). They needed an improved weather and wind prediction model, to reduce the error in their power output prediction.

Hypothesis

  • Studying and analyzing Farm Topology, Hyper Local sensors at wind farms, Other Climate Phenomena.
  • Data from global climatology phenomenon like El Niño (raw weather information at a global level)

Execution

  • Develop a robust forecasting engine using Machine learning techniques on attained data sets.
  • Use it to produce base model that recognizes & evaluates multiple models with hyper local weather sensors and other climate phenomena, delivering both efficiency and higher accuracy in output prediction.

Outcomes

  • Developed forecasting engine with increased accuracy to power generation.
  • 70% reduction in the potential penalty.

Project Highlights

Error rate brought down from

30% to 16%