Employing genetic programming to find the best correlation to predict temperature of solar photovoltaic panels

By A. Sohani, Hoseyn Sayyaadi. 2020

70 citations recorded by the source corpus.

Abstract The best function form to predict the panel’s temperature (Tpanel) is found for a product family of polycrystalline solar panels, with the nominal capacities of 20, 30, 40, 50, 60, 80, 120, 150, 200, 250, 300, and 320 W. For this purpose, genetic programming is used. Experimental data recorded throughout a year is employed while in addition to solar radiation, ambient temperature, and wind velocity, ambient relative humidity is also considered as one effective parameter. First, the best function form is obtained and verified for the 40 W panel, and then, the generalization capability of that is checked for other panels. Moreover, the prediction ability of the best found function form in comparison to the nominal operating cell temperature (NOCT) and nominal module operating temperature (NMOT) approaches, as the most common ways to obtain Tpanel, is evaluated using the monthly and annual profiles of errors. The profiles of error in prediction of Tpanel, efficiency, produced power, and generated energy for the presented, NOCT, and NMOT models are compared together, which shows the vast superiority of the best found function to NOCT and NMOT methods. As an example, for the 50 W panel, the best found function form is able to predict Tpanel, efficiency, produced power, and generated energy 2.15, 3.36, 3.03, and 3.39 times more accurate than NMOT method in a year. It also has 2.82, 4.18, 4.04, and 4.01 times better prediction than the NOCT model during the same period for prediction of the aforementioned performance criteria of the 50 W panel, respectively.

Subfields: Artificial Intelligence, Renewable Energy, Sustainability and the Environment

View Paper: https://www.semanticscholar.org/paper/9155a110d86101ab1cfc0bad0e9205f313d039ea