An Open-Source System Identification Package for Multivariable Processes

We present in this paper an open-source System Identification Package for PYthon (SIPPY¹), which implements different methods to identify linear discrete-time multi-input multi-output systems, in input-output transfer function or state space form. For input-output transfer function models, identification is performed using least-squares regression (FIR and ARX models) or recursive least-squares (ARMAX model). For state space models, various subspace identification algorithms are implemented according to traditional methods (N4SID, MOESP, and CVA) and to parsimonious methods which enforce causal projections. When the model order is not known a priori, three different information criteria can help the user in the choice of the most appropriate order. Many identification and validation tests have been performed on simulation data collected both in open-loop and closed-loop mode. Results show effectiveness and computational efficiency of SIPPY also in comparison with state-of-art proprietary system identification software.