An accurate hybrid intelligent approach for forecasting flicker severity caused by electric arc furnaces

Drastic variations of reactive power consumed by electric arc furnaces (EAFs) often lead to significant voltage fluctuations at the connecting network bus and yield noticeable flickers of lighting devices as well as cause malfunctions of the electrical equipment. If the flicker severity levels are predictable, corrective solutions such as controls of EAF electrodes and reactive power compensators can be developed to mitigate the voltage fluctuations. This paper presents a hybrid approach that combines an improved radial basis function neural network (IRBFNN) and Grey model for the forecast of flicker severity levels. Field measurements are used to train and implement the forecasting model. Test results of ΔV10, short-term flicker severity (Pst) and long-term flicker severity (Plt) obtained by the proposed and five other methods are then under comparisons. Results indicate that more accurate flicker forecast is obtained by adopting the proposed method.