Supervised Learning-Based PV Output Current Modeling: A South Africa Case Study

Photovoltaic (PV) plants utilization for green solar energy is growing exponentially in demand as industries committed to move away from carbon energy sources such as coals, oil, or gas. However, for efficient green solar energy utilization, a precise prediction method is required to minimize design composition wastage. The measured output current determined by empirical method will be compared with the predicted current obtained from the proposed neural network (ANN) and random forest (RF) methods. The comparative analysis of the measured and the proposed models is evaluated by using the minimum root means square error (RMSE), mean absolute percentage error (MAPE), and mean bias error (MBE). The obtained results suggest the superiority of RF over the ANN with improvement performance metrics values of 173% for RMSE, 39% for MAPE, and 188% for MBE.