The prediction of the usual solar irradiation in the Sahel using the artificial neural networks (case study: 50 MW power plant in Nouakchott)

Abstract

The development of a model for predicting meteorological variables using a physical approach was our solution for modeling a solar system, this modeling was carried out in two stages. The first step is to predict the meteorological variable (solar irradiation) at the plant level and the second step is to use a generated energy model to convert these irradiation forecasts into a forecast of the generated energy by the plant. In this study we modeled the solar irradiation curve of the Nouakchott power plant (50 MW) using artificial neural networks (ANN) which create adaptive identification methods and intelligent control laws based on the principal learning, which consists of memorizing previous results and generalizing future results, ultimately modeling the given system. The development of the curve is carried out by carrying out a series of experiments which made it possible to converge towards a methodology offering good precision, using the data measured from solar irradiation over two years at the level of the Nouakchott site. The evaluation of the solar irradiation forecasting model, by calculating the statistical parameters, made it possible to record a normalized average absolute error between 0.121 and 0.126 and a regression factor R (measures the correlation among output-target) with the aid of using 98.4% and 98.5% and the evaluation among specific present techniques in literature display the goodness of the proposed models.