Data-driven identification and model predictive control of biomass gasification process for maximum energy production.

Biomass gasification is an environment-friendly energy conversion process that utilizes bio-waste materials to produce combustible gases. In recent literature, machine learning-based techniques are used to model biomass gasification process. Even though these methods are reported for being viable, developed models' time-independent structure fundamentally limited their prediction capabilities. Furthermore, control of biomass gasification is not studied in the literature despite its importance for industrial applications. We conducted this study in two parts. Firstly, we developed a time-dependent identification model to describe and predict outcomes of biomass gasification using non-linear autoregressive with exogenous neural networks (NARXNN) and experimentally collected data set. The developed model showed exceptional success by reaching R 2 > 0.98 for all output variables. Secondly, we designed a model predictive controller (MPC) in order to control a certain output variable at the desired state. For this purpose, we created polynomial regression models and online optimization routines. Moreover, the designed controller is challenged in practical scenarios such as maximum hydrogen production to test its usability in practical applications. MPC showed satisfactory performance for all scenarios and also showed high compliance with the experimental data which further strengthened its practical usability potential. • First study focuses on the control of biomass gasification process. • Identification is performed by NARXNN with an experimentally collected data set. • NARXNN model predicted all output variables with R 2 > 0.98. • MPC is used to control outputs of the biomass gasification. • MPC applications showed great compliance with the experimental data. [ABSTRACT FROM AUTHOR]