New reactive power calculation method for electric arc furnaces

Improvement of reactive power calculation in electric arc furnaces utilising Kalman filter

“…So, the paid cost for employing KF to reduce the harmonics sensitivity of differential method is gaining an average delay time equal to 1.2 ms. However, the traditional reactive power calculation methods utilised in the SVC control system of EAFs have a delay time about of 10 ms [12]. So, the proposed method with an average delay time equal to 1.2 ms has an acceptable delay in comparison with the traditional methods.…”

“…The main challenge in such control algorithm is reactive power calculation speed and accuracy. Consequently, the performance of the compensation methods is mainly limited by the accuracy and convergence speed of the calculation algorithms [12].…”

“…On the other hand, utilising a method with more delay such as continuous integration in a complete cycle [21] can reduce sensitivity to harmonics. However, differential equation method has the fastest response among the basic methods presented in [12]. It is almost instantaneous but very sensitive to harmonic distortion.…”

Enhancement of SVC performance in electric arc furnace for flicker suppression using a Gray‐ANN based prediction method

“…Collected data are voltage and current samples with sampling time equal to 128 microseconds and length of 100 seconds for each record. By using a method introduced in the previous studies in details, the reactive power corresponding to the fundamental component of voltage and current is attained from voltage and current signals with data windows equal to a full cycle (0.02 s), updated in every half cycle (0.01 s). Hence, each 100‐second reactive power time series consists of 10 000 (=100/0.01) samples.…”

Generalised Cassie–Mayr electric arc furnace models

“…In the second step, ARMA models are utilised for time series modelling. An ARMA model of order ( p, q) which is denoted as ARMA( p, q) is given by z t = w 1 z n−1 + ...+ w p z n−p + a n − u 1 a n−1 − ...− u q a n−q (31) z t is assumed to be a zero average time series. If z t has non-zero average, it can be replaced with z t − μ where μ is the mean of z t .I f in ARMA( p,q) process q=0, it is called AR( p) process and if p=0 it is called MA(q) process.…”