Performance Comparison of ARIMA and k-NN Models for Short-term Wind Speed Forecasting

Abstract

The rapid generation in the wind power is a result of extensive research which is centrally focused on environmental and productive advantages of wind power. A major requirement for this growth is efficient prediction of local wind speed, which provide an estimated potential of wind speed distribution of a region. This paper proposes a methodology for accurate one minute ahead wind speed prediction for two sites of Gujarat region, India. Herein, a comparison between conventional ARIMA and machine learning K-NN model has been presented. Reduced value of performance indices for k-NN makes it a simple and efficient model for prediction of wind speed.