K-nearest neighbors for GEFCom2014 probabilistic wind power forecasting

The paper deals with a forecasting procedure that aims to predict the probabilistic distribution of wind power generation. The k-nearest neighbors algorithm is adapted for this probabilistic forecasting task. It allows quantiles to be estimated without requiring assumptions as to the probability distribution. The influences of several factors (wind speed, wind direction and hour) on the normalized wind power are investigated. The feasibility of the approach is demonstrated through the probabilistic wind power forecasting track of the Global Energy Forecasting Competition 2014.