Neuro-fuzzy modeling and prediction of current total harmonic distortion for high power nonlinear loads

摘要

The paper presents the results of the modeling and the prediction of the total harmonic distortion (THD) of the current and the voltage for a nonlinear high power load. Modeling was performed using intelligent techniques based on neural networks and fuzzy inference. To achieve this, data were measured in the electrical installation of a non-linear electrical load, in this case a high-power electric arc furnace. These data were measured over the entire duration of a steel charge. The measured data was used to train a neuro fuzzy adaptive system. Following training, tests with different architectures have been performed with the neuro fuzzy adaptive system. The modeling and prediction results are useful in designing harmonic current filters. The presence of these harmonic currents decreases productivity and affects the quality of power.