Two-dimensional recursive least squares identification based on local polynomial modeling for batch processes

Abstract

In this paper, a two-dimensional recursive least squares identification method based on local polynomial modeling for batch processes is proposed. Using local polynomial modeling method to parameterize the time-varying characteristics of batch processes, a two-dimensional cost function along both time and batch directions is minimized to design the recursive least squares identification algorithm. Because this proposed method employs local polynomial modeling and utilizes two-dimensional data information to estimate model parameters, it can effectively improve the estimation accuracy and accelerate the convergence rate. Furthermore, the convergence property of the proposed method is analyzed. Finally, the simulation results show the superiority of the proposed method.