Solar Irradiation Components PredictionUsing Meteorological Measures in Odeillo
The rise in the usage of solar energy for electricity generation has led to growing interestamong the researchers to predict the solar radiations at a more granular level. A goodcharacterization of solar irradiation components is also important to produce the best per-formance from photovoltaic plant(PV) and concentrated power plant(CSP). The three mostprominent solar irradiation components include Global Horizontal Irradiation(GHI), DiffuseHorizontal Irradiation (DHI) and Diffuse Normal Irradiation (DNI). The objective of thisproject is to predict the individual components of solar irradiations utilizing added meteor-ological measures. And, also to evaluate the ability of different types of time series modelsin the prediction of individual components of solar irradiation. Solar and meteorologicaldata at the site of Odeillo is used for training and testing the models. Six machine learningmodels namely Classification And Regression Trees, Linear Regression, Stochastic GradientBoosting, KNN, Support Vector Machine and Random Forest model are implemented topredict GHI, DHI and DNI on an hourly basis. Four evaluation metrics were used namelyRMSE,nRMSE, MAE, nMAE. The results concluded that the addition of meteorologicalmeasures significantly helps in improving the prediction of individual solar irradiation com-ponents. This study can help firms dealing with solar installations and energy generationsto decide as to which plant to set up where.