Accuracy assessment of satellite-based and reanalysis solar irradiance data for solar PV output forecasting using SARIMAX

Abstract

Forecasting models are often constrained by data availability, and in forecasting solar photovoltaic (PV) output, the literature suggests that solar irradiance contributes the most to solar PV output. The objective of this study is to identify which between the satellite-based and reanalysis solar irradiance data, namely, short wave radiation (SWR) and surface solar radiation downward (SSRD), respectively, is a better alternative to in situ solar irradiance in forecasting solar PV output should the latter become unavailable. Nine seasonal autoregressive integrated moving average with exogenous variables (SARIMAX) models were presented in this study to assess the forecasting performance of each solar irradiance data together with weather parameters. Using only historical data to forecast solar PV output, three seasonal autoregressive integrated moving average (SARIMA) models were run to forecast solar PV output and to compare and validate the efficacy of the SARIMAX models. The analysis was divided into seasons as defined by the Philippine Atmospheric, Geophysical and Astronomical Services Administration: hot dry, rainy, and cool dry. Results show that the use of SSRD is a better alternative than SWR when forecasting solar PV output for the hot dry season and cool dry season. For the hot dry season, SSRD has an root mean square error (RMSE) value of 0.411 kW while SWR has 0.416 kW. For the cool dry season, SSRD has an RMSE value of 0.457 kW while SWR has 0.471 kW. Meanwhile, SWR outperforms SSRD when forecasting solar PV output during the rainy season, with RMSE values at 0.375 and 0.401 kW, respectively.