Adaptive Forecasting Techniques Applied to Short Time Wind Speed Forecasting

Abstract

Climate change and the increased level of power demand has led to a growing electrical energy production from renewable sources, such as wind power. The main problem associated with wind power production is the nature of the wind speed which is random and non linear. This is a reason why wind speed forecasting is a difficult but crucial task, since its accuracy plays a significant role in achieving reliable and autonomous power production and at the same time contributes in surpassing a series problems, of economic and technical nature. In this study real data is used and the performance of three different techniques for adaptive short term wind speed forecasting are evaluated. The first method combines the multimodel partitioning filter (MMPF) implementing extended Kalman filters (EKF) with Support Vector Machines (SVM), the second is a hybrid method of MMPF and Genetic Algorithms (G.A) and the last method implements an artificial multilayer layer feed-forward neural network (ANN). It should be noted that the first two techniques having the structure presented in this work, have never been tested before on short term wind speed prediction. The results indicate that all three methods are reliable, however the combination of MMPF and SVM provides a more accurate wind speed forecasting. Therefore, the proposed method strengthens the prediction precision, and becomes a significant tool for efficient grid planning.