Deep Learning Based Predictive Compensation of Flicker, Voltage Dips, Harmonics and Interharmonics in Electric Arc Furnaces

In this research work, deep machine learning-based methods together with a novel data augmentation are developed for predicting flicker, voltage dip, harmonics, and interharmonics originating from highly time-varying electric arc furnace (EAF) currents voltage. The aim the prediction is to counteract both response reaction time delays of active power filters (APFs) specifically designed furnaces (EAF). Multiple synchronous reference frame (MSRF) analysis used decompose frequency components EAF current waveforms into dqo components. Then using low-pass future values these components, signals APFs generated. Three different have been developed. two them, Butterworth filter linear finite impulse (FIR)-based or long short-term memory network (LSTM) prediction. third method, convolutional neural (CNN) combined LSTM predict at same time. For 40-ms horizon, proposed provide 2.06%, 0.31%, 0.99% errors prediction, LSTM, CNN respectively. error predicted reconstructed resulted in 8.5%, 1.90%, 3.2% reconstruction abovementioned methods. Finally, Simulink GPU-based implementation predictive APF + trivial 96% 60% efficiency on compensation interharmonics.