Wind Power Prediction using Transformers: A Federated Learning Approach

An increasingly popular renewable energy source is wind power. It aids in overcoming the depletion of non-renewable energy sources and moves a nation closer to sustainable growth. The main challenge of supply and demand optimization in the smart grid is resolved with the aid of wind power prediction, which also lowers operational costs and increases the dependability of the systems. But wind power is mainly distinguished by fluctuations and randomness making it difficult to predict. There are also concerns of privacy due to heavy commercial competition. In this paper, transformers are used for wind power prediction. To overcome the privacy issue, federated learning approach is used to train the transformers. The transformer model trained using federated learning approach is providing the best performance with an RMSE value of 201.25.