Real-Time Detection of Interharmonics and Harmonics of AC Electric Arc Furnaces on GPU Framework

In this paper, a method based on the multiple synchronous reference frame analysis is recommended and implemented to detect time-varying harmonics and interharmonics of rapidly fluctuating asymmetrical industrial loads. The experimental work has been carried out on a typical three-phase alternating current arc furnace installation. In the recommended method, the reference frame is rotated in both directions at speeds corresponding to the positive and negative sequences of all harmonics and all interharmonics 5 Hz apart. To extract the direct current components of the transformed d-q quantities, a low-pass filter is employed. In order to keep the delay of the filter at zero frequency less than a few ms, Kalman estimation technique has been used. Back transformation is then applied for each harmonic and interharmonic component to obtain their positive- and negative-sequences of the associated harmonic and interharmonic in the actual line current waveforms. Parallel computing technique has been applied for the real-time detection of both the phase and the amplitude of all harmonics and interharmonics. This is achieved on NVDIA Jetson TX1 graphics processing unit framework for a sample industrial plant. The developed system is shown to be useful for fast and accurate generation of reference signals for the controllers of the advanced technology power conditioning systems which successfully compensates interharmonics, harmonics, and flicker of the rapidly fluctuating nonlinear industrial loads.