On Computing RMS Indices Through Wavelet Decompositions by Using a Tree with 7 Levels

Abstract

The topology of trees used as support for power quality analysis relying on Discrete Wavelet Transform, Wavelet Package Transform or Stationary Wavelet Transform represents a challenge. Major characteristics which influence the runtime and memory consumption like number of levels and lengths of processed vectors, respectively filters used by the wavelet mother must be considered when certain accuracy is imposed. A tree with 7 levels, 512 components in the root node, used as support for decompositions relying on a Daubechies wavelet mother with filters of 28 components was considered. Root mean square indices for signals acquired with a sampling rate of 35kHz were evaluated. Original calibration techniques were conceived in order to improve the accuracy by using synthetic signals. 3D representations of minimum and maximum percent relative errors yielded by computations are presented and discussed. Simulated and real test signals were used and performance comparisons were performed.