Emerging role of generative AI in renewable energy forecasting and system optimization
Erdiwansyah Erdiwansyah, Rizalman Mamat, Syafrizal Syafrizal, Mohd Fairusham Ghazali, Firdaus Basrawi, S.M. Rosdi
Abstract
• Generative AI improves RES forecasting accuracy by up to 25 %. • GANs and VAEs optimize microgrid and storage operations. • Federated learning enables privacy-preserving energy model training. • Black-box models pose interpretability and regulatory challenges. • AI–IoT convergence supports real-time energy system optimization. The rapid integration of renewable energy sources (RES) into modern power systems introduces significant challenges in forecasting accuracy, grid stability, and energy optimization. Generative Artificial Intelligence (Gen-AI), including architectures such as Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and transformers, offers new capabilities to overcome data sparsity, nonlinearity, and uncertainty in renewable-dominant systems. This study aims to comprehensively review the emerging role of Gen-AI in improving solar and wind forecasting, load prediction, energy storage management, and smart grid optimization. Using a comparative and synthesis-based methodology, this review analyses findings from high-impact publications between 2023 and 2025. Results indicate that GAN-based models reduce root mean square error (RMSE) by 15–20 % in solar irradiance forecasting and significantly enhance spatial-temporal wind simulations. Time-series GAN-LSTM hybrids enhance demand forecasting accuracy under nonlinear conditions, while VAE-driven dispatch models achieve gains of 9–12 % in energy efficiency and curtailment reduction. The novelty of this review lies in mapping Gen-AI's integration with digital twins, federated learning, and AI–IoT frameworks, which enables the real-time, privacy-preserving optimisation of complex energy systems. The principal conclusion is that Gen-AI serves as a transformative tool to enhance system resilience, forecasting precision, and operational flexibility in renewable energy networks. For sustainable implementation, future developments must address challenges in model explainability, data privacy, and scalability. These findings support the journal’s scope by highlighting AI-driven advancements for the reliable, efficient, and sustainable transformation of energy systems.