Nonlinear State Estimation in a Chemical Reactor Using the Ensemble Kalman Filter

This research work presents a comprehensive comparative analysis between two nonlinear filters based on the equations of the discrete Kalman filter, with the aim of improving state estimation accuracy in nonlinear systems. The first filter employs extended linearization techniques to approach optimal estimation, while the second filter utilizes Monte Carlo methods for approximation. The filter equations are developed based on probabilistic principles, providing an approximation to implement the Bayes filter. The study focuses on a mathematical model that represents a chemical reactor utilized in the production of propylene glycol, a crucial substance in the manufacturing of various products, such as creams and pastes. To evaluate the performance of the employed techniques, visual methods and error metrics are used to determine the precision level of the estimators. Through this analysis, the study demonstrates the successful application of the Ensemble Kalman Filter (EnKF) in accurate state estimation, specifically in the context of the continuous stirred reactor. The promising results obtained from the EnKF's application in this study highlight its potential for accurate state estimation in nonlinear systems. Future research can focus on exploring and optimizing the EnKF technique for other complex systems and real-world applications.