Performance of multi-step-ahead-prediction ARX for steam temperature in a self-refilling distillation essential oil extraction system

In this work, performance of multi-step-ahead-prediction (k-SAP) of ARX is evaluated. The data were collected from the self-refilling distillation column using RTD sensor with associated signal conditioning circuit. The control signal is on/off which is excited from pseudo-random binary sequences (PRBS). The PRBS is initially generated using MATLAB with 1 probability band. Then the signal is used by data acquisition application to perturb the system and collect the perturbed system response. The heating system implementing 1.5 kW electrical immersion heater. The power switching is performed using zero-crossing solid-state-relay. Assuming ARX model structure, the input-output data are modeled. Three information criterions have been used to estimate the system model order i.e. normalized sum of squared error (NSSE), Akaike´s information criterion (AIC) and Rissanen´s minimum description length (MDL). In validation stage, the future predictions of these models are evaluated by assessing the k-SAP with k=1 to 100. For each model, the performance indexes i.e. percentages of R² , adjusted R² and NMSE were used.