Dual fluidized bed (DFB) gasification is a promising method for producing valuable gaseous energy carriers from biogenic feedstocks as a substitute for fossil fuels. State-of-the-art DFB gasification plants mainly rely on manual operation or single-input single-output control loops, and scientific contributions only exist for controlling individual process variables. This leaves a research gap in terms of comprehensive control strategies for DFB gasification. To address this gap, we propose a multivariate control strategy that focuses on crucial process variables, such as product gas quantity, gasification temperature, and bed material circulation rate. Our approach utilizes model predictive control (MPC), which enables effective process control while explicitly considering process constraints. A simulation study is given demonstrating how different MPC parametrizations influence the behavior of the closed-loop system. Experimental results from a 100 kW pilot plant at TU Wien demonstrate the successful control achieved by the proposed control algorithm.
In dual fluidized bed (DFB) gasification, the solids circulation rate is critical as it determines the amount of char and heat transported between the interconnected reactors. In DFB plants, multiple control inputs are typically available to control the solids circulation rate, resulting in an over-actuated system. We propose a modeling and control method based on Gaussian process regression, a technique that provides a measure of confidence in the model prediction. The availability of redundant control inputs is resolved by explicitly incorporating the prediction confidence information into the control algorithm to drive the process in regions of low model uncertainty. To address plant-model mismatches, a disturbance model is employed, and an extended Kalman filter is used to estimate both system and disturbance states, enabling offset-free tracking of constant references. Modeling and closed-loop simulation results for both a 100 kW and a 1MW DFB gasification plant demonstrate the applicability of the method to different plants. Experimental results are presented for the 100 kW plant, demonstrating the successful control of the circulation rate by the proposed algorithm.