Multivariable Model Predictive Control to Control Bio-H2 Production from Biomass

Abstract

Two significant units in biomass-based hydrogen plants are the compressor and steam reformer. The compressor works to achieve high pressure for further operations, while the steam reformer produces H2 gas. For the units to operate well against disturbances that may occur (regulatory control) or changes in the set point (servo control), as well as the interactions between the relevant process variables, a Multivariable Model Predictive Control (MMPC) is considered as a controller. The determination of MMPC parameters, including the sampling time (T), prediction horizon (P), and control horizon (M), is crucial for achieving such objectives. Therefore, in this study, MMPC parameter adjustment was performed. The Integral of Absolute Error (IAE) and Integral of Square Error (ISE) were used as control performance indicators. For comparison, we considered the IAE and ISE from the Single-Input Single-Output (SISO)-based Model Predictive Control (MPC) from previous research. As a result, the optimum MMPC parameters were found to be T = 1, P = 341, and M = 121 for the compressor unit, and T = 1, P = 45, and M = 21 for the steam reformer unit. The average increases in control performance (IAE and ISE), compared to the MPC (SISO) used in previous research, were 85.84% for compressor unit 1, 61.39% for compressor unit 2, 94.57% for compressor unit 3, and 73.35% for compressor unit 4. Meanwhile, in the steam reformer unit, the increases in control performance were 63.34% for the heater and 80.16% for the combustor.