Wrapper attribute selector and intelligent systems applied to the identification of residential harmonic sources

Abstract

Signals with a different harmonic content of the fundamental component are present in any power distribution systems due to large quantity and diversity of nonlinear loads connected. In this case, the identification of these loads is necessary in order to mitigate the harmonic currents. Thus, the proposal of this work consists of using an attribute selector (Wrapper) and intelligent systems (Neural Networks and Neural Fuzzy Systems) as an alternative to traditional systems for identification of harmonic sources (nonlinear loads) connected to the electrical system focusing loads commonly encountered in residential customers. Experimental results will be reported and properly justified in order to validate the methodology presented.Index Terms-Harmonic components, intelligent systems, nonlinear loads, identification of harmonic sources.

Linear discriminate analysis and k-nearest neighbor based diagnostic analytic of harmonic source identification

“…Numerous works of literature were stated, that the common machine learning methods, for example, support vector machine (SVM), Naïve Bayes (NB), linear discriminate analysis (LDA), and K-nearest neighbor (KNN) offer satisfactory performance for classifying and diagnosis purpose [46][47][48]. This paper presents high accuracy, fast estimation, and costs effective technique to diagnose the type of harmonic sources in the distribution system with single-point measurement at the point of common coupling (PCC) by utilizing the machine learning algorithms [49,50].…”

A Diagnostic Analytics of Harmonic Source Signature Recognition by Using Periodogram

“…Unfortunately, the parameter of the phase angle of the currents can compromise the harmonic source identification as shown by [12]. As a result, it is proposed in [13] a research on the method to identify harmonic sources using the amplitude of the measured current only. An attribute selector named Wrapper is combined with the DFT of current for the data pre-processing step to obtain the most relevant amplitudes of harmonic current orders.…”