Online PV Monitoring and Prediction Using Tree-Based Method

Online monitoring and prediction play a crucial role in ensuring the optimal performance of grid-connected photovoltaic (PV) stations. On the other hand, using machine learning techniques, specifically tree-based methods, has demonstrated its effectiveness in predicting and detecting faults in nonlinear processes. In this study, we employed a tree regression method for the online monitoring and prediction of a 7 kWp PV station located in the desert region of Adrar. Our approach exhibited clear superiority when compared to classical regression methods such as the Linear Least Squares Method (LLS). We achieved a determination coefficient
$(\mathbf{R}^{\mathbf{2}})$
of 0.9671 and a Mean Absolute Error (MAE) of 187.2, surpassing the results obtained with LLS, which yielded an
$\mathbf{R}^{\mathbf{2}}$
of 0.9646 and an MAE of 252.06, highlighting the efficacy of our proposed methodology.