Multiple harmonic-source classification using a Self-Organization Feature Map network with voltage–current wavelet transformation patterns

Abstract

a b s t r a c tThis paper proposes multiple harmonic-source classification using a Self-Organization Feature Map (SOFM) network with voltage (V)-current (I) wavelet transformation patterns. Using the V-I wavelet transformation (WT) patterns, a SOFM network is employed to separate non-harmonic loads from non-linear loads in a distribution system. Morlet wavelet functions are used as feature extractors to extract the features from voltage and current signals. These features are constructed from various V-I WT patterns, and can vary with different dilation and translation parameters. Therefore, the selectable features are relatively broad for real-time applications. Two-dimensional (2-D) patterns appear in different harmonic fluctuations with various harmonic orders, load levels, and power factors. A SOFM network is employed to classify the various V-I WT patterns, including non-linear electronic devices, AC/DC motors, and Electric Arc Furnaces (EAFs). By contrast with the traditional SOFM network and the support vector machine (SVM), the testing results show that the proposed method has a fast learning process, a high accuracy, and an adaptive capability with new add-in training patterns. It can be used as an added tool for power quality (PQ) engineers and can be integrated into monitor instruments.

Identification of multiple harmonic sources in power system containing inverter‐based distribution generations using empirical mode decomposition

“…The number of train and test samples for any classes are 100 samples. As mentioned before, connected harmonic sources contain four NLs (6-pulse and 12-pulse rectifiers, EAF, and DC motor) [6,8,15], three inverter-based DGs (DFIG, PV, MT) and one SVC. Wind turbines using DFIG consist of a wound rotor induction generator and an AC/DC/AC IGBT-based PWM converter modelled by voltage sources.…”

“…On the other hand, to identify the type of multiple harmonic sources at the PCC, the single point methods are preferred to the multiple point methods. Therefore, several single point methods have been proposed for identification of multiple harmonic sources, which use the direction of harmonic powers [10], the Norton equivalent circuit [11][12][13], and neural networks [14][15][16]. The success of total harmonic powers approach using several comprehensive case studies was investigated by Omran et al [10].…”

“…Chang et al [14] proposed a radial-basis-function-based (RBF) neural network for identification of harmonic sources at the PCC. In [15,16], different neural networks such as self-organisation feature map network, feed forward back propagation neural network (FFBPN), and support vector machine (SVM) using waveforms of voltages and currents have been used for identification of multiple harmonic sources at the PCC. The main disadvantage of these neural networks is that they are timeconsuming.…”

A Critical Review of Time-frequency Distribution Analysis for Detection and Classification of Harmonic Signal in Power Distribution System

“…To deal with non-stationary signal, short-time Fourier transform (STFT) is introduced [14], [15]. Nevertheless, STFT is not proficient to detect the dynamic signal property because the window width fixed to affect the frequency-time resolution [6], [8], [16]. Whereas, the time-frequency resolution depends upon the size of the window [17], [5].…”

“…Furthermore, the ST is a reliable technique to characterize the harmonic parameters [26], [28], [22]. The ST also offers an extraordinary multiresolution analysis while characterizing the harmonic components [9], [18], [16], [22]. Based on above discussion, an evaluation and a critical review of TFDs for harmonic signal detection and classification are crucial and need to be done.…”