Closed-loop and open-loop identification of an industrial wastewater reactor

We benchmark eight different system identification algorithms on an industrial wastewater reactor under both closed loop and open loop conditions. The algorithms studied in the paper include subspace identification, partial least squares, ordinary least squares, and output error approach. The studied model structures include ARX, ARMAX, and FIR models. The minimum description length principle and Akaike information criterion are used to select the order of these models. Comparison of the eight different algorithms are given based on open loop experimental data and closed loop operational data from a real industrial wastewater reactor.