State estimation of a chemical reactor process model - A comparative study

Abstract

Due to the challenges associated with measuring some of the key variables of chemical processes, state estimators are often used to overcome this problem. This paper deals with the problem of state estimation of a chemical process model representing a continuously stirred tank reactor (CSTR) using the Extended Kalman Filter (EKF), Particle Filter (PF), and recently developed Variational Bayesian Filter (VBF). The VBF has been recently proposed to solve the nonlinear estimation problem because it can be applied to large parameter spaces, has better convergence properties and relatively easy to implement. Here, a comparative study is conducted to compare the estimation performances of these three estimation techniques in estimating the two states (the concentration and temperature) of the CSTR process model. Simulation results show that the VBF has improved state estimation performance over both EKF and PF, and the PF shows improved state estimation performance over EKF.