Titel
State Estimation of UV Flash Processes
Abstract
State estimation is concerned with reconstructing the state of a process based on a model and on noisy measurements. State estimation is necessary in model predictive control, model identification (e.g. parameter estimation), monitoring, and prediction.
Models of UV flash processes combine the second law of thermodynamics with the principles of mass and energy conservation. The UV flash is relevant to rigorous models of chemical phase equilibrium processes such as flash separators, distillation columns, and oil reservoir production.
We consider four different types of state estimation algorithms, also called filters: 1) the extended Kalman filter, 2) the unscented Kalman filter, 3) a particle filter, and 4) the ensemble Kalman filter. The filters update the state estimates when new measurements become available, and they use simulations to predict the process behavior in between measurements. We give a qualitative introduction to the four filters and we discuss the differences between them in terms of estimation accuracy and computational performance.
We present a numerical example where the states of a UV flash separation process are estimated based on temperature and pressure measurements. We compare the accuracy and the computational efficiency of the filters.