Joint dynamic data reconciliation/parameter estimation: Application to an industrial pyrolysis reactor

Highlights: • Data reconciliation/parameter estimation in a pyrolysis reactor is analysed. • Kalman filter and quasi-steady-state assumption are incorporated. • A multi-channel scheme is assessed for overcoming the loss of observability. • Single and multi-channels approaches are tested. -- Abstract: The monitoring and control of thermal cracking processes are critical to maintain a stable and safe operation. This contribution presents alternatives for joint dynamic data reconciliation and parameter estimation with application to a pyrolysis reactor. A decoupled filtering approach based on the quasi-steady-state (QSS) assumption is assessed for simultaneous estimation of state variables and unknown reactor parameters. Two different state/parameter decoupling procedures are proposed, the single and multi-channel schemes. The multi-channel approach is designed for overcoming the loss of observability due to the large number of states and parameters to be estimated. The performances of both strategies are contrasted and evaluated using the operational data of an industrial pyrolysis reactor. Although both approaches depict a good performance, the multi-channel technique presents a higher speed of convergence and a broader capability in the determination of model parameters.