Application of Deep Learning Algorithms for Scenario Analysis of Renewable Energy-Integrated Power Systems: A Critical Review

As the global shift towards renewable energy sources accelerates, the challenge of effectively modeling the inherent uncertainty associated with these energy units becomes increasingly significant. Sustainable energy sources, like solar and wind power sources, are highly variable and difficult to predict, making their integration into power systems complex. Beyond renewable energy, other critical sources of uncertainty also influence power systems’ operation, including fluctuations in electricity prices and variations in load demand. To address these uncertainties, stochastic programming has become a widely adopted approach. Preparation of the required scenarios for a stochastic programming framework typically includes two main components: scenario generation and reduction. Scenario generation involves creating a diverse set of possible future outcomes based on various uncertainties considered, while scenario reduction focuses on refining these scenarios to a manageable number without losing an essential piece of information. In this paper, we explore the innovative methods used for scenario generation and scenario reduction, with a special emphasis on deep learning approaches. Additionally, we provide future research recommendation, identify areas for further development, and discuss the challenges associated with these deep learning methods.