Constrained two dimensional recursive least squares model identification for batch processes

Recursive system identification, due to its easy online implementation and computation efficiency, has been widely used in many advanced process controls such as adaptive control and model predictive control (MPC). This paper proposes a novel two dimensional recursive least squares identification method with soft constraint (2D-CRLS) for batch processes. This method can improve the identification performance by exploiting information not only from time direction within a batch but also along batches. A soft constraint term is incorporated in the cost function to reduce the variation of the estimated parameters. A bound on weighting matrix has been established as the sufficient consistency condition in the paper together with a practical guideline for weights selection. Results based on the experimented data for injection molding, show the superiority of the proposed method over the conventional identification based on recursive least squares.