A Survey on Wind Speed Forecasting in Smart Grids Using Deep Learning Algorithm Applications

The crisis resulting from the use of fossil fuels has heightened the push towards increasing the share of renewable energy production, particularly from sources like wind and solar power. While wind energy holds special significance due to its potential availability round the clock, relying on wind power comes with various challenges. One such challenge is the uncertainty surrounding the production levels of wind farms. In response to this challenge, researchers have consistently sought ways to predict wind speed, with one of the most crucial methods being the application of machine learning techniques. These data-driven methods exhibit a commendable ability to forecast wind speed by leveraging historical data and other relevant inputs. Recent years have witnessed substantial advancements in deep learning methods, further enhancing the appeal of these approaches. To achieve this objective, this article endeavors to explore the role of applying deep learning methods in wind speed forecasting. A comprehensive survey of the application of these algorithms is conducted, and ultimately, a comparative analysis of these methods is performed, taking into account the specific types of wind speed forecasting problems. To achieve this, the article delves into the examination of nine machine learning methods, namely convolutional neural network (CNN), recurrent neural network (RNN), long short-term memory (LSTM), deep belief network (DBN), physics-informed neural networks (PINN), Autoencoder, generative adversarial network (GAN), deep reinforcement learning (DRL), and hybrid models. The goal is to provide insights into the effectiveness of these approaches and to facilitate a comprehensive understanding of their application in addressing wind speed forecasting challenges.