A statistical model for computing global spectral solar irradiance under all-sky conditions at four stations situated in main regions of Thailand

Abstract

This paper presents a statistical model for computing global spectral solar irradiance under all-sky conditions at four stations situated in main regions of Thailand. The model expresses the global spectral irradiance under all-sky conditions as a multiplication of two functions, namely, a function of global spectral irradiance under clear-sky conditions and a cloud modification function. To develop the model, global spectral solar irradiance was measured at four stations in the main regions of Thailand, namely Chiang Mai (18.77˚ N, 98.97˚ E) in the northern region of Thailand, Ubon Ratchathani (15.25˚ N, 104.87˚ E) in the north-eastern region of the country, Nakhon Pathom (13.82˚ N, 100.04˚ E) in the central region of this country, and Songkhla (7.18˚ N, 100.60˚ E) in the southern region of Thailand. The spectral data from these stations were gathered and divided into two groups. The first group (January 2017 to December 2020) was used for modeling and the second group (January 2021 to December 2021) for validating the model. The first function was developed using the first group of data. Clear-sky conditions were detected using sky images obtained from a sky camera installed at each station. To develop the second function, a satellite-derived cloud index derived from Himawari-8 data at the stations was utilized. To validate the model of the global spectral solar irradiance under all-sky conditions, the model was employed to compute the global solar irradiance at the four stations for the year 2021, and the findings were compared to the global solar irradiance determined from the measurements taken at the four stations. It was discovered that the spectral irradiance calculated from the model and that acquired from the measurements was in good agreement. There was a discrepancy in the form of root mean square difference relative to the mean measured value (RMSD) of 9.48%. We concluded that the model performed well in computing the global spectral solar irradiance.