Solar radiation estimation models based on artificial intelligence applied to the photovoltaic electrical generation for norte de Santander, Colombia

Abstract

The absence of direct measurements of solar radiation in many countries around the world (due to the high costs of installation and maintenance of the measuring devices) has been identified (in the Colombian case by the Energy Mining Planning Unit) as one of the main barriers for the deployment of photovoltaic systems. In this sense, estimation techniques have been developed in the literature for locations where this variable is not measured. These techniques take advantage of the correlation between the irradiance and other climatic parameters of wider distribution and easier access to construct models that forecast with high accuracy the solar potential of a specific place. Thus, the current research exposes the implementation of an indirect estimation model designed with Artificial Intelligence that uses the temperature, humidity, wind speed and sunshine duration to predict the irradiation, as a tool for sizing photovoltaic systems in Norte de Santander (Colombia) where solar radiation measurements are available just in three of the 40 municipalities in which the region is geographically divided.