Identification and direct digital control of an electric arc furnace controller

A method for the identification of an accurate system model and subsequent digital control of an electric arc furnace electrode position controller is presented. The identification procedure involves model reduction to derive a low-order model from a theoretically realised higher-order system model, together with parameter tuning in the low-order model using an extended Kalman filter applied to input-output data collected from a real controller perturbed by a pseudorandom binary sequence. Such tuning is necessary to correct for inaccuracies in the initial theoretical system modelling. The technique produces an accurate system model suitable for designing digital control algorithms but avoids the computational difficulties associated with other identification techniques which treat the system as a black box. The paper concludes with consideration of the forms of digital control suitable for the system.