Analysis of ARMA Solar Forecasting Models Using Ground Measurements and Satellite Images

As the solar photovoltaic (PV) share in the electricity grid is growing year by year, solar irradiance forecasting is becoming increasingly important. In this work the performance of a recursive formulation of ARMA models suitable for operational context using the Pampa Húmeda region as a case study is analyzed. Results are promising, as this simple adaptive algorithm does not require historical data and outperform persistence at all lead times. The improvement produced by adding satellite cloudiness data and short-term local variability as exogenous inputs is also evaluated. It is found that the spatially averaged satellite albedo is a useful input variable, improving the forecast performance, while the introduction of short-term variability produce negligible performance changes under this kind of models.