Effect of Noises on the Performance of Nonlinear State Estimator for Semi Batch Autocatalytic Esterification Reactor

To enhance the product quality of sec-butyl propionate ester in a semi batch reactor, on-line dynamic optimization and control can be implemented. However, to implement the control scheme, online measurement is an issue that needs to be solved. The most common way is to apply an estimator such as the extended Kalman filter (EKF) to estimate the required variables. Unscented Kalman filter (UKF) can also be implemented as an alternative derivative free stochastic estimator. In this work, the performance of UKF and EKF is tested on various scenarios of Gaussian (G) and non-Gaussian (NG) state and measurement noise sequences that occur in a semi batch autocatalytic esterification reactor to produce sec-butyl propionate ester. From the results, the UKF is found to outperform the EKF in most cases. Meanwhile, the EKF has superior performance over the UKF if no noise or low levels of noises occur. It is also observed that the UKF gives optimal filter for all levels of G noises but poses a limited level in compensating the NG noises.