Did you know that seasons affect solar PV output forecasting accuracy?
Literature suggests that mean ambient temperature, solar irradiation and cloudiness, atmospheric particles, wind speed, precipitation, and extreme weather affect solar PV production output [1]. This implies that forecasting models are dependent on geography and season. And from our recent paper, we looked into the solar PV output forecasting accuracy based on various seasons defined by the Philippine Atmospheric, Geophysical, and Astronomical Services Administration (PAGASA), using the production data from a solar PV installation in Saint Louis University, Baguio City.
PAGASA [2] characterized the seasons in the country into two, mainly wet (June to November) and dry (December to May) seasons, each lasts for six months. Meanwhile, the dry season can be subdivided into cool dry (December to February) and hot dry (March to May). Based on these classifications, we used the Seasonal Autoregressive Integrated Moving Average models with Exogenous variables (SARIMAX) to forecast solar PV output.
Results of our study suggest that hours N-1 and N-24 help predict hour N for the hot dry season; hours N-1, N-24, and N-48 help predict hour N for the cool dry season; and hours N-1, N-2, and N-24 help predict hour N for the rainy season.
Meanwhile, the cool dry season forecasting model shows higher accuracy (in terms of Mean Absolute Error) followed by the hot dry season, then the rainy season, at 2.26, 4.82, and 12.91, respectively.
You can read more about our ICUE paper here: http://bit.ly/SINAG22ICUE
References: