0093-9994 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2019.2911568, IEEE Transactions on Industry Applications Analysis of Power System Harmonic Subgroups of the Electric Arc Furnace Currents Based on a Hybrid Time-Frequency Analysis Method Erhan Sezgin, Student Member, IEEE, Özgül Salor, Senior Member, IEEE Gazi University, Electrical and Electronics Engineering, Ankara, Turkey sezginerhan@gazi.edu.tr, salordurna@gazi.edu.tr Abstract—In this paper, a hybrid time-frequency analysis method, specially developed to decompose harmonic subgroups and interharmonics of the electric arc furnace (EAF) currents, which are highly time varying due to the operation principles of the EAFs, is presented. The main objective is accurate perception of harmonics and interharmonics in cases of rapid changes or power quality (PQ) events in power system voltages and currents. Harmonic and interharmonics detection has been achieved using Discrete Wavelet Transform (DWT), which provides time-localization in cases of highly time-varying signals. Although DWT elicits accurate spectral decomposition at low frequencies, and especially at the baseband, bandwidths of the band-pass filters increase which results in loss of accuracy at higher frequencies. In order to avoid this problem, power signals are modulated by complex exponential waveforms, which corresponds to shifting the required harmonic sub-band contents to the baseband, where the accuracy of the DWT is the best. Using the proposed hybrid combination of DWT and complex exponential modulation, time domain waveform of each harmonic sub-band of the EAF currents can be estimated close to ideal values. The method also enables to focus on any required interharmonic sub-band in addition to the harmonics. To optimize the performance of DWT in cases of PQ events, various windowing approaches are discussed. The proposed harmonic and interharmonic estimation method has been verified on both synthetic data and EAF currents collected from the electricity transmission system. Estimated harmonic and interharmonic waveforms can serve as a good reference in many areas including active power filtering operations, limit violation determinations, and etc. in the power system. Index Terms — Complex Exponential Modulation, Discrete Wavelet Transform, Electric Arc Furnace (EAF), Frequency Shifting, Interharmonics, Power System Harmonics, Wavelets. I. INTRODUCTION Electrical Arc Furnaces (EAFs) are special types of loads for a power system, being the sources of unexpected power system frequency components, such as even harmonics and significant amounts of interharmonics in addition to the expected power system harmonics. Moreover, these components are highly time-varying, i.e. the amplitudes of the components need to be recomputed in almost every cycle of the fundamental frequency. Therefore, analysis of EAF currents and voltages requires special attention. Power system harmonics are usually analyzed by Discrete Fourier Transform (DFT) based frequency analysis methods [1]. The calculated DFT coefficients have no time dependence and the variation of the components in time cannot be traced, once the DFT coefficients are obtained, which is an important disadvantage when time-varying EAF currents are considered. Short Time Fourier Transform (STFT), which is the DFT applied on short-time windowed signals, is introduced to keep track of the variation of the frequency components in time [1]. In the DFT-based methods, in order to increase the resolution in the frequency-domain, the window size in time-domain should be increased, which causes the loss of localization in time, i.e. the variations of the frequency components in time are observed as their averages inside the window. For the signal to be stationary inside the analysis window, the window size has to be reduced, however, this causes DFT frequency resolution to be decreased. As an alternative to DFT-based methods for the harmonic analysis in power systems, Discrete Wavelet Transform (DWT) has been proposed to overcome the time-frequency resolution problem [2]. DWT is implemented by applying filter banks on signals, which make it possible to separate frequency bands from each other. Decomposition is achieved in two consecutive steps, which are High Pass (HP) filtering and Low Pass (LP) filtering first and then down sampling operations to produce detail and approximation coefficients, respectively [3]. The same procedure is applied on the approximation coefficients at each turn repetitively to obtain increased frequency resolution. As the coefficients calculated by DWT stand for the signal contents bounded in the frequency bands, time domain values of power, energy or RMS can be computed directly using these coefficients with proper mathematical operations [4]. Similar approach has been used for harmonic analysis of power signals in various research work in the literature [4] - [5]. Dyadic distribution of the sub-bands in DWT analysis may not be efficient for the power signal analysis, where uniform harmonic subgroup computations are required. Filtering high-frequency coefficients together with the low- frequency coefficients at each level repetitively will yield a uniform frequency subset contrary to DWT and this operation is called as Discrete Wavelet Packet Transform (DWPT). The coefficients obtained by DWPT have also