Solar irradiance forecast using aerosols measurements: A data driven approach

Abstract

The use of renewable energy resources has grown several fold in the last two decades. One of the main challenges is the uncertainty in their output power due to fluctuating meteorological conditions like sunshine intensity, cloud cover and humidity. In desert areas, another parameter that has a significant impact on solar irradiance is dust, which has been neglected in many studies. In this work, an hour-ahead solar irradiance forecasting model is proposed, this model utilizes both Aerosol Optical Depth (AOD) and the Angstrom Exponent data observed from a ground station at the previous hour. The proposed model was tested under different widely used data driven forecasting models, including Multilayer Perceptron (MLP), Support Vector Regression (SVR), k-nearest neighbors (kNN) and decision tree regression. Applying the MLP model using data from Saudi Arabia shows a root mean square average error of under 4% and forecast skill of over 42% for one-hour ahead forecast. The proposed forecasting model demonstrates a superior accuracy compared to other models when tested and verified under different feature selection schemes. The MLP model is especially applicable for desert areas under clear sky conditions, where dust storms are frequent and AOD in the air is high (>0.4).