An artificial neural network to predict solar UV radiation in Tabriz

Abstract

Introduction: Solar radiation has a major role in design, utilization, development, and planning of solar energy. The most important source of natural ultraviolet radiation is the sun, which has an important role in many biologic processes. Some of these processes are useful, like the production of vitamin D in the body, or curing rickets, and some of them are not, such as skin inflammation, premature aging, and eye diseases like cornea inflammation and cataracts. Because of lack of important information about the amount of ultraviolet exposure in most cities and weather stations, using methods based on artificial intelligence has been suggested. This study has been conducted to evaluate artificial neural networks ability to predict ultraviolet exposure based on experimental data. Materials and Methods: Firstly, the amount of ultraviolet radiation types A, B and C have been measured for a whole year from sunrise to sunset in Tabriz during 2016-2017. To apply the ANN in current study, there are six neurons in the input layer corresponding to the input data (UVA, UVB, UVC, visible light intensity, month of year and hours of day), one hidden layer with three neurons was identified through a preliminary trial-and-error, and one neuron in the output layer for simulate and prediction of solar ultraviolet exposure. Two statistical indexes, RMSE and R2, have been used to evaluate the offered model. Results: The predicted results using the artificial neural network in this study, showed that ANN advanced model able to forecast solar ultraviolet exposure, according to error metrics. Average errors obtained for simulation was RMSE=0.0001 with R2=0.98. Conclusion: The results showed that developed ANN model is capable of simulating the amount of solar ultraviolet exposure.