Employing Best Input SVR Robust Lost Function with Nature-Inspired Metaheuristics in Wind Speed Energy Forecasting
Wind power has been experiencing a quick improvement. Without a doubt, wind is a variable asset that is hard to forecast. For instance, traditionally time series, extra holds are distributed to deal with this uncertainty. This paper presents a comparison of the performance of various Support Vector Regression (SVR) applied to short-term wind power forecasting. The analogy with BORUTA and multivariate adaptive regression splines (MARS) as judge best input and employ genetic algorithm and particle swarm optimization to find best parameter in Support Vector Regression with robust lost function. We measure the accuracy of this models by Symmetric means absolute percentage error (sMAPE) and we get the best model BORUTA-SVR-PSO with sMAPE 2.07155%. Moreover, we measure the energy conversion using Feedback Linearization Control (FLC).
IAENG International Journal of Computer Science
Rezzy Eko Caraka, Rung Ching Chen, Sakhinah Abu Bakar, Muhammad Tahmid, Toni Toharudin, Bens Pardamean, and Su Wen Huang
