| Abstrakt: |
Sustainability of the earth depends on renewable energy. Forecasting the output of renewable energy has a big impact on how we operate and manage our power networks. To guarantee grid dependability and permanence and to lower the danger and expense of the energy market and systems, accurate forecasting of renewable energy output is essential. Researchers have been drawn to this topic by deep learning's recent success in a variety of applications, and its bright future is shown in the variety of proposed approaches and the rising number of publications. This paper reviews deep learning-based solar and wind energy forecasting research that has been published over the past five years, covering in-depth topics such as the data and datasets used in the reviewed works, data pre-processing techniques, deterministic and probabilistic techniques, and evaluation and comparison methods. To facilitate methodological comparisons, the key traits of all the reviewed publications are compiled in tabular form. The field's existing difficulties and potential areas for future research are described. According to trends, recurrent neural network models, including those with long short-term memories and gated recurrent units, are second most frequently utilized in this sector behind hybrid forecasting models, with convolutional neural networks coming in third. We also observe an increase in interest in probabilistic and multistep forecasting techniques. Using the important learnings from this thorough study, we also create a comprehensive taxonomy of the research, which will, in our opinion, be essential for comprehending the cutting-edge and promoting innovation in this area. [ABSTRACT FROM AUTHOR] |
