Incorporation of parametric uncertainty into complex kinetic mechanisms: Application to hydrogen oxidation in supercritical water

Abstract In this study, uncertainty analysis is applied to a supercritical water hydrogen oxidation mechanism to determine the effect of uncertainties in reaction rate constants and species thermochemistry on predicted species concentrations. Forward rate constants and species thermochemistry are assumed to be the sole contributors to uncertainty in the reaction model with all other model parameters and inputs treated as deterministic quantities. The analysis is conducted by treating the model parameters as random variables, assigning each a suitable probability density function, and propagating the parametric uncertainties through to the predicted species concentrations. Uncertainty propagation is performed using traditional Monte Carlo (MC) simulation and a new, more computationally efficient, probabilistic collocation method called the Deterministic Equivalent Modeling Method (DEMM). Both methods predict virtually identical probability distributions for the resulting species concentrations as a function of time, with DEMM requiring approximately two orders of magnitude less computation time than the corresponding MC simulation. The results of both analyses show that there is considerable uncertainty in all predicted species concentrations. The predicted H 2 and O 2 concentrations vary ± 70% from their median values. Similarly, the HO 2 concentration ranges of +90 to −70% of its median, while the H 2 O 2 concentration varies by + 180 to − 80%. In addition, the DEMM methodology identified two key model parameters, the standard-state heat of formation of HO 2 radical and the forward rate constant for H 2 O 2 dissociation, as the largest contributors to the uncertainty in the predicted hydrogen and oxygen species concentrations. The analyses further show that the change in model predictions due to the inclusion of real-gas effects, which are potentially important for SCWO process modeling, is small relative to the uncertainty introduced by the model parameters themselves.