On the role of lagged exogenous variables and spatio–temporal correlations in improving the accuracy of solar forecasting methods

Abstract

a b s t r a c tWe propose and analyze a spatioetemporal correlation method to improve forecast performance of solar irradiance using gridded satellite-derived global horizontal irradiance (GHI) data. Forecast models are developed for seven locations in California to predict 1-h averaged GHI 1, 2 and 3 h ahead of time. The seven locations were chosen to represent a diverse set of maritime, mediterranean, arid and semi-arid micro-climates. Ground stations from the California Irrigation Management Information System were used to obtain solar irradiance time-series from the points of interest. In this method, firstly, we define areas with the highest correlated time-series between the satellite-derived data and the ground data. Secondly, we select satellite-derived data from these regions as exogenous variables to several forecast models (linear models, Artificial Neural Networks, Support Vector Regression) to predict GHI at the seven locations. The results show that using linear forecasting models and a genetic algorithm to optimize the selection of multiple time-lagged exogenous variables results in significant forecasting improvements over other benchmark models.

Combining solar irradiance measurements, satellite-derived data and a numerical weather prediction model to improve intra-day solar forecasting

“…Therefore, due to the restriction of only one year, we took the decision to select three weeks of a month for training and one week for testing purposes since this study only possess one year available. In this manner we can assure that training and testing datasets share similarities regarding meteorological behaviors, while keeping a ratio of 75% training and 25% testing [28].…”

“…We considered a maximum number of 30 satellite derived radiation data based on the calculation explained by Mazorra et al [30]. In order to study weather conditions in Gran Canaria Island and establish the best satellite pixels between whole gridded data at the different stations, we computed the Pearson correlation between satellite and ground data [27,28]. The huge amount of satellite derived data makes the computation difficult, so we applied a median filter for each 3 Â 3 satellite pixels.…”

“…As explained in Ref. [30], the distribution of pixels in the different time-lagged images is according to the number of correlation values over 0.5 in each station [28]. To decide the best selection of pixel, we computed six different tests during the training process.…”

“…As explained in Ref. [30], the distribution of pixels in the different time-lagged images is according to the number of correlation values over 0.5 in each station [28]. For each station, four tests were calculated using the annual correlation results and two more test used a different distribution every quarterly group of images.…”

“…In the last years, several studies used different techniques such as autoregressive models (AR) [26e28], artificial neural networks [29,30] and genetic algorithms for choosing the relevant information from satellite-derived data [28].…”