Identifying statistical properties of solar radiation models by using information criteria

• Analyzing ground measurement data to improve solar irradiance modeling.
• Identifying probabilistic distributions in two models using information criteria.
• Gaussian classical distribution is unsuitable for solar irradiance modeling.
• Suitable statistical distributions laws are proposed.
• Procedure generates solar irradiance data statistically comparable to measured data.

The purpose of this article is to improve modeling of statistical properties of solar radiation models through the analysis of measurement data on the ground in the intertropical zone. For this, we identify, using information criteria, the probabilistic distributions introduced in two models of synthetic solar radiation generation. We then validate the results by using the KL divergence and KSI parameter as comparison criteria between the distributions arising from real and synthesized data. Our study confirms, for example, that the Gaussian classical distribution is not suitable for modeling solar radiation, and we propose other more suitable statistical laws instead. The value of the identification procedure of the distribution laws presented in this article is that it ensures the generation of solar radiation data comparable in their statistical content to the measured data. Another advantage is that this procedure contributes to highlighting the time invariance of distribution laws representing the random terms. We conclude that this new information-criteria-based method permits the identification of the probability laws that best describe the statistical distributions introduced in the models of synthetic solar radiation generation.