Short Term Power Forecasting of Wind Energy System using Deep Neural Network (DNN) Trained by a Novel Meta-Heuristic Optimization Algorithm

Stochastic nature of wind power with high amount of non-linearity makes it very difficult to predict wind power production in real time, which has a high impact in renewable energy industry. The uncertainty of wind power makes it challenging to integrate it with the power grid. As a solution, an early short term forecasting of the wind power significantly improves the wind power generation. For this purpose, a novel Archimedes optimization algorithm (AOA) is used to train a deep neural network (DNN) for short-term wind power prediction. Effective exploration and exploitation behavior of AOA for updating the particles position, effectively trains the deep neural network. To validate the performance of the proposed technique, well-known methods are compared using case studies. The proposed method has shown better prediction performance as compared to existing techniques and achieves up to 96.7% and 98.4% less training error and up to 96.6% and 97% less testing error in winter and summer seasons respectively.