Diminishing the Computational Burden when Analyzing Electrical Signals with Long Wavelet Filters

Journal Title: ANALELE UNIVERSITATII DIN CRAIOVA - Seria Inginerie electrica - Year 2017, Vol 41, Issue 41

Abstract

Wavelet analysis is characterized by filters with irrational coefficients and runtimes increasing with filters order. The terminal nodes of Wavelet Packet decomposition trees exhibit good harmonics’ selectivity when implemented with Daubechies wavelet mothers with filters of length 28 and respectively of length 40 (denoted in MATLAB by ‘db14’ and ‘db20’). Therefore they can be used for power quality (PQ) analysis of electric signals. When rounded to 4 decimal places, certain coefficients from these filters become 0. In the “rounded” context, the number of iterations sup-porting the convolution products involved by the Wavelet analysis is restricted. Based on the number of neglected co-efficients and length of wavelet filter, a formula is deduced for the percent runtime savings strictly associated to the wavelet decompositions. Statistical evaluations (considering sets of 20 consecutive programs executions) were performed considering full operational contexts, where all types of studied wavelets were involved in complex operations like evaluating PQ indices and harmonic spectra. The mean runtime savings proved to be 20.4% for ‘db14’ and respec-tively 23.2% for ‘db20’. The impact of rounding over the PQ analysis’s accuracy was evaluated considering three-phase waveforms acquired from a power group. The follow-ing quantities were considered: harmonic spectra and ener-gies of the Wavelet Packet tree’s terminal nodes, instanta-neous values corresponding to the fundamental frequency, vectors of details and approximations yielded by the Dis-crete Wavelet Transform and signals reconstructed from the vectors of details and approximations. Highly acceptable small errors were obtained, recommending the rounding to 4 decimal places of the analyzed wavelet filters as a faster reliable wavelet based analysis method.

Authors and Affiliations

ILeana-Diana Nicolae, Petre-Marian Nicolae, Ionuţ Daniel Smărăndescu

Keywords

Related Articles

Digital Control for the Air Pressure with Multi-Characteristics Selection

The paper deals with a preparatory research stage for controlling the air pressure inside a small tank by various characteristics and algorithms. Basic elements both for the hardware design and for the software support a...

 Modern Techniques for Monitoring Circuit Breakers by using Microcontroller System

 The paper deals with experimental monitoring system, developed by the authors, by using a microcontroller within the 8051 family, Dallas DS87C550. This micro-system has two main parts: the acquiring and transfer mo...

AN AGGREGATE MULTI-LOOP CONTROL FOR A DC SERVODRIVE. THE DESIGN OF THE NEURO-FUZZY CONTROLLER

The purpose of the study is the synthesis of a reliable and robust fuzzy controller for a 2nd order DC servodrive. A first step is to prepare the computations associated to the main control loops for the above mentioned...

MONITORING SYSTEM FOR THE LOCAL LAUNCH POINTS OF THE ANTI-HAIL UNITS

This paper presents a system for monitoring an anti-hail system. It is proposed a specific system for anti-hail network in Romania witch completes the known systems with information subsystems specific for both to identi...

EXPERT SYSTEM FOR POWER RECTIFIERS DIAGNOSIS

In designing of the static power converters, the experience of an expert is absolutely necessary in order to obtain a competitive solution. The evolution of the CAD methods has brought in the last time expert systems w...

Download PDF file
  • EP ID EP264024
  • DOI -
  • Views 110
  • Downloads 0

How To Cite

ILeana-Diana Nicolae, Petre-Marian Nicolae, Ionuţ Daniel Smărăndescu (2017). Diminishing the Computational Burden when Analyzing Electrical Signals with Long Wavelet Filters. ANALELE UNIVERSITATII DIN CRAIOVA - Seria Inginerie electrica, 41(41), 34-39. https://europub.co.uk/articles/-A-264024