A review of deep learning transformer variants used in solar PV forecasting

Abstract

Solar power has become one of the most important renewable energy sources as it is sustainable, its prices are reducing, and it could contribute to the transition to clean energy across the world. Nevertheless, the inherently intermittent nature of solar PV output due to the rapidly fluctuating weather conditions demands a very serious need to have accurate forecasting systems. The current reviews are primarily over traditional deep learning or statistical models, creating a knowledge gap on how transformer-based architectures intend to separate long-term dependencies and nonlinear behaviours of solar PV forecasting. Consequently, the objective of this review is to fill this gap by giving a detailed analysis of the variants of transformers that would be utilized when predicting PV power. A large variety of transformer models, among which are Simple Transformer, Informer, Temporal Fusion Transformer (TFT), Patch-Transformers, TSformer, and Hybrid Transformer models, among other recent models, are systematically analysed in this paper. Their architectural innovations, approach enhancement, benefits, and drawbacks in the forecasting of solar energy are described in the review. The strategy adopted also involves comparative appraisal, architecture-based classification, and analysis based on published benchmark investigations in terms of performance metrics. The result of this review demonstrates that transformer-based models are better than classic AI algorithms to capture long-range temporal patterns, process heterogeneous inputs, and achieve better forecasting results. The findings of the literature show further that strong increases in robustness, interpretability, and predictability of multiple horizons are achieved. Altogether, this review provides a unified picture of the modern transformer-based developments and defines the future research that ought to facilitate the improvement of scalability, uncertainty quantification, and the deployment of PV forecasting systems in the real world.