Multi-Model Predictive Control of a novel Rapid Pressure Swing Adsorption system

A Multi-Model Predictive Control (M-MPC) algorithm is developed for a novel Rapid Pressure Swing Adsorption (RSPSA) oxygen concentrator system. The RPSA uses a zeolite material to purify oxygen from ambient feed air, and was designed to produce ~ 90% oxygen for medicinal therapies. A standard MPC algorithm was implement on the RPSA which rejects disturbances and improves set point tracking in the 90% operating range, but a RPSA can produce a wide range of oxygen purities. The proposed M-MPC algorithm uses a collection of linear models each of which corresponds to a different oxygen purity range. The linear models are identified around different operating points, and use Pseudo-Random Binary Signal (PRBS) simulation data in the identification process. With this improved control algorithm, the RPSA can produce a range of oxygen purities, and reject typical RPSA process disturbances. M-MPC switching rules, operating region boundary considerations and set point tracking cases are presented and discussed.