Evaluating the effect of particulate matter pollution on estimation of daily global solar radiation using artificial neural network modeling based on meteorological data

Abstract

Nowadays, using solar energy as a clean, renewable and available energy resource, has increasingly become crucial due to dwindling of fossil fuel resources and many environmental concerns across the world. Thus, numerous research works have been conducted to study the implementation of this energy source in different industries as a replacement for current, less environmentally-friendly energy sources. Engineering designs and scientific studies for improving the efficiency of solar energy systems and assessing their feasibility require actual data regarding the amount of solar radiation, which cannot easily be measured in some regions. For these locations, empirical relations and theoretical models are common methods of estimating the amount of solar irradiance. Recent studies have illustrated that empirical methods and models are helpful in forecasting the amount of solar radiation and despite their acceptable accuracy, using novel methods such as Artificial Neural Network (ANN) modeling enables researchers to make even more accurate predictions. For estimating daily GSR, importance of key input parameters including Particulate Matter (PM), in addition to widely used modeling parameters in recent studies, such as Relative Humidity, Wind Speed and Daily Temperature has been assessed. Also, based on meteorological data, an ANN modeling approach has been developed for a 1-year period and for Tehran, Iran as a case study location. Some statistical indexes such as Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE) and Absolute Fraction of Variance (R2), which are mainly used for evaluating the efficiency and accuracy of models, have been determined as 1.5%, 0.05 J Cm−2d−1 and 99% respectively. It can be asserted that based on the values of the aforementioned statistical indexes, presented model in this research has relatively higher accuracy and credibility compared to the models that have been presented before. In addition, this paper shows that considering Particulate Matter (PM) as a parameter in estimation of GSR boosts modeling efficiency.