Identification of state models using principal components analysis

A predictive state-space dynamic plant is identified using a two-stage approach based on principal components analysis. The procedure is applied to a simulated benchmark problem known as the overheads condensor reflux drum (OCRD) model, a non-linear multivariable plant with mixed dynamics. The identified model is validated against an independent test set and its step and frequency responses compared with a linearised analytical model of the OCRD.