Prediction of the Performance of a Sun Tracking Photovoltaic System using different Artificial Intelligence Techniques: Case Study in Zarqa, Jordan

Abstract

The article proposes the use of Artificial Intelligence (AI) models to predict the performance of a sun tracking Photovoltaic (PV) system built-in Zarqa City, Jordan. The system is off-grid with various Azimuth angles and tilt angles. The study involved taking various measurements over a 5-month period. The prediction models employed Artificial Neural Networks (ANN) with five different prediction classifiers, namely, random forest, forest tree, multilayer perceptron (MLP), BPF regression, and linear regression, to predict the performance of the sun-tracking PV system using experimental data. Different metrics are used to demonstrate and validate the accuracy of the proposed models. It is found that all proposed prediction models are of great accuracy. The best prediction classifier is found to be a forest tree classifier with an R2 value of 99.79% and a minimum absolute relative error of 2.36%. Moreover, the least accurate prediction classifier is found to be the linear regression with an R2 of 95.27% and an absolute relative error of 25.71 %.