Prediction of electrical energy generation from photovoltaic plants with NARX neural network

The integration of renewable energy sources into existing power grids presents significant challenges due to their inherent variability. This research investigates the application of Nonlinear Autoregressive Exogenous (NARX) neural networks for accurate prediction of electrical energy generation from photovoltaic plants. Solar energy forecasting is crucial for effective grid management, energy trading, and optimal operation of photovoltaic installations. The study employed a correlational research design to analyze the relationship between NARX neural networks (variable X) and the accuracy of electrical energy generation prediction from photovoltaic plants (variable Y). Using time series data from two operational solar plants—a 112.04 kW installation in La Palma del Condado (Huelva, Spain) and a 75.24 kW facility in Alcalá del Río (Seville, Spain)—the research developed predictive models that incorporated both historical energy production values and solar radiation measurements. The methodology followed a systematic approach including neural network construction, training, validation, and testing phases. Performance evaluation through linear regression analysis revealed remarkably high prediction accuracy, with correlation coefficients (r) of 0.991 and 0.968 for the two case studies, respectively, yielding an average accuracy of 97.9%. The results demonstrate that NARX neural networks provide highly reliable forecasting capabilities for photovoltaic energy generation, which can significantly contribute to reducing operational costs, minimizing the need for backup energy sources, and facilitating the broader integration of solar power into existing electricity grids.