Vertical extrapolation of wind speed using artificial neural network hybrid system

Abstract

Different approaches have been used for the extrapolation of wind speed to the turbine hub height which are mainly based on logarithmic law, power law and various modifications of the two. This paper proposes two artificial neural network (ANN) hybrid system-based models using genetic algorithm and particle swarm optimization, namely GA-NN and PSO-NN for vertical extrapolation of wind speed. These models are very simple in a sense that they do not require any parametric estimation like wind shear coefficient, roughness length or atmospheric stability. Rather they use available measured wind speeds at 10, 20 and 30 m heights to estimate wind speed at higher heights up to 100 m. Proposed methods have been compared with ANN, power law and logarithmic law. Daily, monthly and yearly average values at different heights were investigated by proposed models. Predicted values at 30 and 40 m heights were compared with actual measured wind speeds. In every investigation, the mean absolute percentage error and coefficient of determination values were found to be less than 5 % and more than 0.98, respectively. Comparatively low values of mean square error of around 0.05 were also observed while comparing with other existing methods. Although GA-NN and PSO-NN have almost similar performance, GA-NN was found to be performing little better than PSO-NN.