Use of support vector regression and numerically predicted cloudiness to forecast power output of a photovoltaic power plant in Kitakyushu, Japan

The development of a methodology to forecast accurately the power produced by photovoltaic systems can be an important tool for the dissemination and integration of such systems on the public electricity grids. Thus, the objective of this study was to forecast the power production of a 1-MW photovoltaic power plant in Kitakyushu, Japan, using a new methodology based on support vector machines and on the use of several numerically predicted weather variables, including cloudiness. Hourly forecasts of the power produced for 1?year were carried out. Moreover, the effect of the use of numerically predicted cloudiness on the quality of the forecasts was also investigated. The forecasts of power production obtained with the proposed methodology had a root mean square error of 0.0948?MW?h and a mean absolute error of 0.058?MW?h. It was also found that the forecast and measured values of power production had a good level of correlation varying from 0.8 to 0.88 according to the season of the year. Finally, the use of numerically predicted cloudiness had an important role in the accuracy of the forecasts, and when cloudiness was not used, the root mean square error of the forecasts increased more than 32%, and the mean absolute error increased more than 42%. Copyright (c) 2011 John Wiley & Sons, Ltd.