Publication

Abstract : A technique of feature extraction used for wind speed, power planning of small grids with hybrid power sources. In the past many techniques, i.e., SVD, FFT and DWT tried for feature extraction and cluster formation. The second-generation wavelet transforms, i.e., Lifting wavelet transform has been in use in diverse application, with much better performance. Motivation to carry out the present work intended for implementation of the algorithm of feature extraction and cluster formation in real time application. The technique applied based upon random number theory with central mean theorem and inequalities to find centroids of clusters of wavelet coefficients for feature extraction. The algorithm implementation carried out on data taken from National Energy for Research Laboratory. The datasets are spread across a period of one year. Results demonstrate the recovery of maximum energy of the signal with minimum error as compared to original data at the third level. It justifies the selection of third level of resolution, aimed at maximum signal energy recovery from the featured data and comparison of statistical results.