A Modified Moving Horizon Estimation Scheme for Multi-timescale Chemical Processes

Many chemical processes have timescale multiplicity arising from the coupling between different physico-chemical phenomena. A direct application of classical moving horizon estimation (MHE) to multi-timescale processes may require a long estimation horizon, leading to a high computational load. In this work, we explore a modified MHE scheme where output measurements are selectively chosen based on process dynamics to be used in the MHE optimization problem. This allows the MHE to cover measurements in the distant past up until the latest available measurements without significantly increasing the computational complexity. Simulation studies have shown that the proposed approach provides similar accuracy of state estimation compared to classical MHE with a significant reduction in computational time.