A comparative study of optimal hybrid methods for wind power prediction in wind farm of Alberta, Canada

In the recent years, by rapid growth of wind power generation in addition to its high penetration in power systems, the wind power prediction has been known as an important research issue. Wind power has a complicated dynamic for modeling and prediction. In this paper, different hybrid prediction models based on neural networks trained by various optimization approaches are examined to forecast the wind power time series from Alberta, Canada. At first, time series analysis is performed based on recurrence plots and correlation analysis to select the proper input sets for the forecasting models. Next, a comparative study is carried out among neural networks trained by imperialist competitive algorithm (ICA), genetic algorithm (GA), and particle swarm optimization approach. The simulation results are representative of the out-performance of ICA in tuning the neural network for wind power forecasting.