A New Dynamic Intelligent Time Domain Arc Furnace Modeling based on Combination Adaptive Neuro-fuzzy Inference System and Chain Code

Abstract

Electric arc furnaces have intense non-linear, time variant, and uncertain characteristics bringing about voltage and current harmonics, unbalances, and voltage flicker. In this article, a new three-phase dynamic and intelligent arc furnace in time domain based on adaptive neuro-fuzzy inference system is proposed. In this approach, the adaptive neuro-fuzzy inference system is trained by several actual electric arc furnace operations, and a pattern detection based on the modified chain code method is presented. Hence, electric arc furnace performance is analyzed by applying a preprocessing stage using low-pass average filters and chain code method that recognizes the defects in different patterns such as pulses, impulses, similar and inverse profiles, steps, etc. A chain code is a lossless compression algorithm for detecting the behavior of different data. The adaptive neuro-fuzzy inference system is simultaneously trained by electric arc furnace operation at each stage by means of the patterns. For data collection, a power quality analyzer and oscilloscope are connected to different actual electric arc furnaces. Finally, the electric arc furnace model is presented as a voltage source depending on a current similar to a non-linear black box. This model has no linearity in arc characteristic and can show an actual arc.