Wind Speed Prediction Based on Chaos Theory using RBF Neural Networks

Wind speed prediction can be regarded as significant factor in control of wind turbines, schedule of the connection/disconnection of turbines and stability guarantee of power grids which is commonly carried out in various approaches. In this paper, a chaos based approach by analyzing only the previous measured data is proposed. For this purpose, in addition of evaluating the chaotic nature of wind speed data, the chaos theory with Neural Network techniques in forecasting session are combined in order that we can propose a method for wind speed prediction. For this regard, at first the correlation dimension and largest lyapunov exponent of wind speed time series are computed to prove that wind data generator process is chaotic. Then phase space of data generator dynamic is reconstructed. In this regard, we use the False Nearest Neighbors (FNN) algorithm to determine the embedding dimension and Average Mutual Information (AMI) approach to measure time delay for phase space reconstruction. Afterwards, Multi Layers Perceptron (MLP) neural networks and Radial Basis Function (RBF) neural networks are proposed to predict the wind speed which its structure is designed based on time delay and embedding dimension data. At the end, proposed methods apply on real data and results are expressed.