The rapid integration of renewable energy sources such as wind and solar has introduced unprecedented levels of uncertainty, nonlinearity, and multi-scale coupling into modern energy systems, challenging conventional forecasting and control methodologies. While recent advances in physics-informed learning and large-scale artificial intelligence models have independently demonstrated promising results, existing research remains fragmented across forecasting, control, and system modeling domains. Most prior reviews focus either on small-scale physics-informed models or on data-driven large models without addressing their integration into closed-loop renewable energy operation. This paper identifies a critical gap: the absence of a unified framework that systematically connects physics-informed large models with renewable energy forecasting and control across multiple temporal and spatial scales. To address this gap, we propose a novel Physics-Informed Large-Model Framework for Renewable Energy Forecasting and Control (PILM-REC) that integrates physical knowledge embedding, large-model representation learning, uncertainty-aware forecasting, and decision-oriented control within a single conceptual architecture. This review synthesizes recent advances in physics-informed neural networks, neural operators, transformer-based spatiotemporal models, and hybrid learning–control approaches, mapping them onto the proposed framework. By structuring the literature through the lens of the PILM-REC framework, this work provides new insights into current limitations, emerging opportunities, and future research directions toward scalable, interpretable, and reliable large-model-driven renewable energy systems.
Sulaima et al. (Mon,) studied this question.