Hybrid methodology for modeling short-term wind power generation using conditional Kernel density estimation and singular spectrum analysis

A fundamental part of the probabilistic forecasting of wind energy process is to take into account wind speed forecasts. To achieve accurate probabilistic forecast of wind output, it is developed a hybrid methodology using a nonparametric techniques known as SSA (Singular Spectrum Analysis) and (CKD...

Full description

Saved in:
Bibliographic Details
Published in:Dyna (Medellín, Colombia) Colombia), 2017, Vol.84 (201), p.145-154
Main Authors: Aguilar-Vargas, Soraida, Castro-Souza, Reinaldo, Pessanha, José Francisco, Cyrino-Oliveira, Fernando Luiz
Format: Article
Language:English
Subjects:
CKDE
Density
Economic forecasting
Electric power generation
ENGINEERING, MULTIDISCIPLINARY
estimación condicional de la densidad por kernel
forecasting
Generación de energía eólica
Lead time
previsión
Random variables
Renewable energy
series temporales
Spectrum analysis
SSA
time series
Wind power
Wind power generation
Wind speed
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:A fundamental part of the probabilistic forecasting of wind energy process is to take into account wind speed forecasts. To achieve accurate probabilistic forecast of wind output, it is developed a hybrid methodology using a nonparametric techniques known as SSA (Singular Spectrum Analysis) and (CKDE) Conditional Kernel Density Estimation. SSA is employed to forecast wind speed and CKDE to obtain probabilistic forecasts of wind energy, based on the fact that wind power generation has a nonlinear relation with the wind speed and both are random variables distributed according to a joint density function. A Brazilian hourly wind dataset including wind speed and wind power is used to illustrate the approach. Once the wind speed forecasts are obtained the corresponding probabilistic forecast of the wind power generation is estimated for a lead time of 24 hours ahead. The results obtained are compared with other existing methodologies.
ISSN:0012-7353
2346-2183
DOI:10.15446/dyna.v84n201.59541