A Predictive Reactive Power Measuring Based on Time Series and DLSL Algorithm for Compensating Applications

Electric reactive power is one of the effective control parameters in the static Var compensators of power systems. These compensators have a time delay due to their natural performance such as thyristor ignition and reactive power measurements. Therefore, in this paper, a new predicted reactive meter based on time series and delta least squares lattice (DLSL) is proposed to avoid the mentioned time delay. At first, the electric arc furnace (EAF) based on the hidden Markov model and actual measured data are modeled. Then, reactive power model coefficients of time series are adequately calculated based on the actual voltage and current data collected from modeling ac EAFs as time-varying and unbalanced load that can generate voltage and current harmonics, voltage flicker, and reactive power destruction in the main busbar. Now, to determine the prediction relationship coefficients online, the DLSL algorithm that has a very good performance for fast sampling is applied. Finally, this method is implemented in the control system of thyristor-controlled reactor/fix capacitor connected on the same busbar. The experimental results based on actual measuring data show the accuracy and validity of the presented method.