Enhancing Renewable Energy Forecasting: A Novel Approach Integrating Satellite Imagery with Deep Learning Models

One of the priorities in addressing the serious challenge of integrating renewable energy into the power system is to improve the accuracy and robustness of forecasts for renewable energy generation. Deep learning models, utilizing historical power generation and meteorological forecasts, have become central in predicting renewable energy outputs. Based on that, this study introduces an innovative approach of incorporating satellite cloud imagery as an alternative, robust data source for forecasting. This paper proposed a novel framework integrating three types of data: historical power generation, traditional meteorological forecasts, and image-based satellite data. The integration not only improves forecast precision but also fortifies the prediction system against data manipulation and external threats. Numerical studies reveal that this approach not only addresses the technical needs of accurately predicting power generation in an increasingly renewable-dependent energy system but also fortifies the forecasting process against potential vulnerabilities, ensuring a more stable and secure transition towards sustainable energy solutions.