Development of Hybrid Model based on Artificial Intelligence for Maximizing Solar Energy Yield

Nedeljko DUČIĆ¹, Snežana DRAGIĆEVIĆ¹, Pavle STEPANIĆ², Nebojša STANKOVIĆ¹*, Milan MARJANOVIĆ¹
¹ University of Kragujevac, Faculty of Technical Sciences, 65 Svetog Save, 32000 Čačak, Serbia
nedeljko.ducic@ftn.kg.ac.rs, snezana.dragicevic@ftn.kg.ac.rs,
nebojsa.stankovic@ftn.kg.ac.rs (*Corresponding author), milan.marjanovic@ftn.kg.ac.rs
² Research and Development Institute Lola L.T.D., 70A Kneza Viseslava Street, 11000 Belgrade, Serbia
pavle.stepanic@li.rs

Abstract: This paper presents an approach that utilises artificial intelligence to maximise the potential of solar energy. The proposed method involves a hybrid intelligent system that combines machine learning and genetic algorithms. The primary objective is to optimise the solar energy yield by accurately estimating global solar radiation on a horizontal surface. Four different machine learning techniques are employed for the estimation of global solar radiation. These techniques utilise measured data obtained from a local weather station to develop predictive models. The most accurate model is selected to design the fitness function for optimising the tilt angle of the solar collector. The aim is to maximise energy yield by determining the optimal angle using a genetic algorithm. Results show that the proposed model effectively identifies the daily optimal angle for maximising solar radiation on the tilted surface. The developed model demonstrates an increase in global solar radiation on a tilted surface compared to the optimal angle model commonly used in practical applications.

Keywords: Machine learning, Genetic algorithm, Hybrid system, Measured data, Solar radiation prediction, Optimal tilt angle.

>>FULL TEXT: PDF

CITE THIS PAPER AS:
Nedeljko DUČIĆ, Snežana DRAGIĆEVIĆ, Pavle STEPANIĆ, Nebojša STANKOVIĆ, Milan MARJANOVIĆ, Development of Hybrid Model based on Artificial Intelligence for Maximizing Solar Energy Yield, Studies in Informatics and Control, ISSN 1220-1766, vol. 32(4), pp. 95-104, 2023. https://doi.org/10.24846/v32i4y202309