Wind field reconstruction using non-negative matrix factorization and principal component analysis with CFD basis

For the purpose of selecting the best sites for installation of wind farms and for increasing the net yield of wind energy, the wind speed is required to be determined at different positions, within a domain of interest. This helps to determine the natural variance/uncertainty in the wind speed, which is very useful for predicting the wind power potential in an area. For this, a huge amount of data is required to be processed (from wind speed sensor measurements) and mathematical algorithms are required for rapid reconstruction of the wind field. The Non-negative Matrix Factorization (NMF) and Principal Component Analysis (PCA) have been presented, which can be applied for reconstructing the wind field around obstruction models using CFD basis. The absolute reconstruction error tends to increase with an increase in the inlet velocity. The relative accuracy of NMF and PCA are subject to the sampling rate of the measurement, but there is no influence on the distribution of the wind speed sensors around the obstructions model (above a sampling rate of 0.05%). By application of these reconstruction models using WSR, it has been concluded that the NMF and PCA can be adequately used to reconstruct the wind field around an obstruction model.