The Wind Speed Forecasting is crucial in the feasibility analysis of potential Wind Turbine site. This paper presents a comparative analysis of nonlinear autoregressive (NAR) neural network model and nonlinear autoregressive with exogenous inputs (NARX) neural network model for forecasting wind speed from short term data set. The neural network models are trained and tested with Levenberg Marquardt, Scaled Conjugate Gradient and Bayesian Regularization algorithms. This work uses the LIDAR wind speed data collected at Gulf of Khambat, off Gujarat coast, installed by National Institute of Wind Energy, Chennai. Various architectures of NAR and NARX models are implemented and the performances are compared. The results revealed that the NARX model trained with Levenberg Marquardt algorithm outperformed in comparison with NAR model for short term wind speed forecasting.
Volume 12 | 04-Special Issue
Pages: 1907-1912
DOI: 10.5373/JARDCS/V12SP4/20201679