Robust Data Reconciliation Applied to Steady State Model with Uncertainty

Abstract

Different researchers have proposed the treatment of uncertainties in measurements because they interfere in the process state estimation. Data reconciliation procedure improves the information supplied by measurements, minimizing the discrepancy existent between measurements and accurate process model. This problem allows obtaining unbiased estimation when measurements follow exactly a normal distribution. Nevertheless, the presence of outliers do not allow the use of the former procedure, therefore Robust Data Reconciliation is developed. This latter provides accurate solutions when measurements follow approximately the normal distribution. Although many advances have been developed to treat measurement uncertainties in Data Reconciliation framework, there are not research works that consider model and measurement uncertainties simultaneously in presence of outliers. In this work, a Simple robust Method, which takes advantage of temporal redundancy, is applied to benchmarks that contain uncertain parameters. Performances measures are tested for different magnitudes of simulated outliers and compared with the ones provided by a classic Data Reconciliation procedure. Results show that the Robust Data Reconciliation procedure can yield unbiased estimations of measurements and parameters when outliers and parametric uncertainties are present.