SYSTEM-IDENTIFICATION OF A 3-PHASE SUBMERGED-ARC FERROSILICON FURNACE

A 36MW submerged-are ferrosilicon (FeSi) furnace is modeled using the technique of system identification, The main purpose of the modeling is for the simulation and evaluation of different control schemes, In particular, the part of the system associated with the control of the three-phase electrode currents, which is done by positioning of the electrodes, is examined, Data were collected at Icelandic Alloys Ltd., using experiments conducted mostly in open loop, i,e,, data for the dynamic system identification using electrode positions as inputs and electrode currents as outputs, as well as data for the estimation of the effect of conductance changes, electrode-to-hearth voltages, and other disturbances, using fixed electrode positions and electrode currents as outputs, The positioning of electrodes responding to control signals was estimated and modeled using measurements in closed loop, It was decided to model the process dynamics and the disturbance environment separately, since data with inputs varied as well as fixed were readily available and since the disturbances are slowly varying compared to the process electrode position/current dynamics, Data collection for the disturbance identification was done by keeping the three electrodes fixed and recording the currents, A first order AR (autoregressive). model was used to model the disturbance environment, Data collection for the dynamic system identification was done by varying the three electrode positions and recording the electrode positions and the currents, Different ARX (autoregressive with auxiliary input) MIMO (multiple-input/multiple-output) models were tried using several model orders, combinations of inputs, and input delays, Then,dependent parameters due to symmetry in the three electrodes, as well as common factors (zeros) in the three input (electrode) polynomials, were eliminated, This was done by using the indirect prediction error method (IPEM) [1], which numerically minimizes a cost function of the difference between the dependent parameters and the new (independent) ones, Effectively, the results obtained in this study indicate that the simple linear models developed include all critical factors needed in a simulation of different electrode position based control schemes for the electrode current control of the furnace,