Harvesting Solar Energy: Prediction of Daily Global Horizontal Irradiance Using Artificial Neural Networks and Assessment of Electrical Energy of Photovoltaic at North Eastern Ethiopia

Abstract

The difficulty and high price of measuring devices make the utilization of solar energy impractical, particularly in developing countries like Ethiopia. Because of its variability and nonlinear characteristics, it needs accurate prediction techniques in a specific location. Thus, the objectives of this article were: (i) assessing daily global horizontal irradiance using network types-activation functions of artificial neural network (ANN); and (ii) evaluating the daily energy delivered to and available on photovoltaic (PV) cells of GaAs at Kemissie, Woldiya, and Hayk, in the northeastern part of Ethiopia. Nine parameters were used in the input layer, and daily GHI was the output result. Feed forward back propagation (FFBP) and cascade forward back propagation (CFBP) with tansig, logsig, and purelin of ANNs were used. The best pairs were FFBP-logsig, CFBP-logsig, and CFBP-tangsig, with 0.8882 ≤ r ≤ 0.9833, respectively. The average values were (4.374 kWh/m²/day ≤ GHI ≤ 6.805 kWh/m²/day) at Kemissie, (4.246 kWh/m²/day ≤ GHI ≤ 7.116 kWh/m²/day) at Hayk, and (4.479 kWh/m²/day ≤ GHI ≤ 7.011 kWh/m²/day) at Woldiya. The energy delivered to and obtainable from PV cells varied between 0.1274 and 0.2135 kWh and 0.1101 and 0.1844 kWh, respectively, for all sites. This bears out the suitability of the site for the installation of a solar energy system.