Tree-based ensemble methods for predicting the module temperature of a grid-tied photovoltaic system in the desert

ABSTRACT

The PV module temperature is a crucial parameter in the performance of a grid-tied PV station and it has an important effect on the PV system efficiency. In this work, we are interested in predicting the module temperature of a grid-tied photovoltaic system using tree-based ensemble methods, namely random forest and boosted decision tree. The linear least square method and the artificial neural network method were used as a frame of reference to evaluate the results of tree-based ensemble methods. The hyper-tuning of the tree ensemble method was done to optimize the model’s parameters and to improve accuracy and prevent overfitting. All developed models have similar accuracy during the training and they are equally applicable for predicting PV module temperature. The results showed that during testing, the tree-based ensemble methods maintained their accuracy with R2 above 0.98. Meanwhile, the accuracy of other methods declined, which proves the utility of the tree-based ensemble over the classical method especially the ANN

KEYWORDS:

• Photovoltaic
• module temperature
• machine learning
• prediction

Acknowledgments

The authors are very grateful to the Directorate General for Scientific Research and Technological Development, URERMS/CDER (Algeria) to have provided the necessary funding and support to realize this project.