Model-based multivariable control of dual fluidized bed gasification

The dual fluidized bed (DFB) gasification process is a promising technology for producingvaluable energy carriers from biogenic raw materials as a substitute for fossil fuels. Currently, DFB gasification plants are predominantly operated manually or with basic single-input control loops, while comprehensive scientific approaches for controlling key process variables are lacking.This dissertation aims to address this research gap by developing, implementing, and experimentally validating innovative control strategies. The focus is on controlling criticalprocess variables, such as the gasification temperature, product gas mass flow, and solidscirculation rate. This is intended to significantly increase the level of automation and enhancethe competitiveness of the technology. The control concepts developed were tested on a 100 kW pilot plant at TU Wien. The development of advanced control methods is based on mathematical models that describe the relationships between control inputs, disturbances, and output variables. The first part of the work focuses on identifying these models, employing both physical and data-driven approaches. Where feasible, relationships were represented by physical equations, such as mass and energy balances or thermodynamic equilibria. Complementary to this, data-driven methods like artificial neural networks or Gaussian process regression were applied to specific aspects, such as solids circulation rate. Model parameters were derived from measurementdata gathered from both previous experiments and dedicated identification tests conducted onthe 100 kW pilot plant.A central contribution of the dissertation is the development of a model predictive controller (MPC) to control key process variables like gasification temperature and product gas mass flow. The MPC accounts for numerous constraints, including a required residual oxygenconcentration in the exhaust gas and minimum gas flow rates necessary for reactor fluidization.Additionally, a subordinate controller controls the solids circulation rate by adjusting airflow in the combustion reactor.The final part of the dissertation examines the control of solids circulation rate in greater detail, presenting alternative approaches for controlling this variable. These include a linear minimum variance MPC and a nonlinear controller based on Gaussian process regression. Since data-driven models provide varying prediction accuracy across different operating conditions, the developed control strategies leverage information on model accuracy and redundant actuators to guide the process in areas with higher model reliability. The nonlinear approach offers a flexible and plant-independent solution, validated through simulations onthe 100 kW pilot plant and a 1 MW demonstration facility. It was subsequently implemented and successfully tested on the 100 kW pilot plant. Both the simulation and experimental results demonstrate that the presented methods can successfully control key variables in the DFB gasification process, thereby addressing the research gap in control strategies for such plants.