SE2GPA stacked ensemble framework for solar photovoltaic energy generation prediction and alerting system

Abstract It has become increasingly challenging for electric utility providers to meet energy demands due to the rise in the use of energy-hungry devices. Therefore, in order to produce energy at our end, we must use renewable energy resources like solar energy. To anticipate the generation of solar photovoltaic energy, this paper proposed the Stacked Ensemble framework for Solar Photovoltaic Energy Generation Prediction and Alerting System (SE2GPA). The fundamental models used in this proposed stack-based ensemble learning approach include Gradient Boosting, Extreme Gradient Boosting, LightGBM, and CatBoost. To provide even more precise results, this paper used Cat Boosting as the meta-model and result shows the better performance by reducing error of energy prediction compared to state-of-art papers. The proposed model achieved an RMSE of 2.02 and an R² of 0.68, outperforming existing state-of-the-art models. Furthermore, this paper employed an alerting system to alert the user if the efficiency of the solar photovoltaic system drops significantly below some threshold value. The model was tested on a single dataset, which may restrict its generalizability. Furthermore, expanding the framework to multiple datasets, integrating real-time streaming data, and enhancing the alerting mechanism with predictive maintenance features.