Regional forecasts and smoothing effect of photovoltaic power generation in Japan: An approach with principal component analysis sciencedirect/science/article/pii/S0960148114000986 Renewable Energy Volume 68, August 2014, Pages 403–413 Joao Gari da Silva Fonseca Juniora, Corresponding author contact information, E-mail the corresponding author, E-mail the corresponding author, Takashi Oozekia, Hideaki Ohtakea, Ken-ichi Shimosea, Takumi Takashimaa, Kazuhiko Ogimotob Highlights • We proposed a method to obtain regional forecasts of photovoltaic power. • Its accuracy was compared with persistence and with a no PCA approach. • Principal component analysis use was important to yield good regional forecasts. • The effect of principal component analysis was comparable to the regional smoothing effect. • The method outperformed a persistence method by at least 51% regarding the rmse. Abstract Regional forecasts of power generated by photovoltaic systems have an important role helping power utilities to manage grids with a high level of penetration of such systems. The objective of this study is to propose a method to obtain one-day ahead hourly regional forecasts of photovoltaic power when regional information is available. The method is based on the use of principal component analysis, support vector regression and weather forecast data. One-day ahead regional forecasts of photovoltaic power were done for 4 of the main regions of Japan for 1 year, 2009, using hourly power generation data of 453 photovoltaic systems. The performance of the method was characterized comparing the results it yielded with the ones provides by a persistence approach and by an approach that do not employ the principal component analysis. Moreover, the expected smoothing effect on the error achieved when the regional forecasts are based on forecasts for each photovoltaic system is presented, constituting an additional reference to evaluate the proposed method. The results show that the method performed well; its regional forecasts had a normalized annual root mean square error of 0.07 kWh/kWrated in the worst case, and the persistence approach was outperformed by at least 51% regarding the same error. The use of principal component proved to be a simple and particularly effective approach, decreasing the bias of the forecasts in all regions, and causing a reduction of the normalized root mean square error from 20.2% to 57.8% depending on the region. The proposed method also yielded results within the same level of forecasts which benefitted from the smoothing effect; the former presented a maximum variation of 10.2% of the normalized root mean square error of the latter in the worst case. Keywords Photovoltaic systems; One-day-ahead regional forecasts; Smoothing effect; Principal component analysis; Support vector regression
Posted on: Wed, 19 Mar 2014 06:15:38 +0000
Trending Topics
Recently Viewed Topics
© 2015