Property Uncertainty Analysis and Methods for Optimal Working Fluids of Thermodynamic Cycles

There is an increasing interest in recovering industrial waste heat at low tempera-tures (70-250◦C). Thermodynamic cycles, such as heat pumps or organic Rankine cycles, can recover this heat and transfer it to other process streams or convert it into electricity. The working fluid, circulating around the cycle, is vital for the per-formance of the cycle. Computational modelling of working fluid properties and cycle processes allows to identify promising working fluid candidates together with optimal cycle conditions.
However, such computer simulations are subject to modelling uncertainties due to the operational conditions, process correlations and fluid properties. In this thesis the focus lies on the uncertainties from physical and chemical property data, caused by the experimental measurements or by the prediction models.
This thesis project presents a comprehensive framework to assess property un-certainties for different levels of thermodynamic cycle models. The framework con-sists of 1) a methodology for the development and uncertainty analysis of group contribution based property models, 2) a Bootstrap method for the quantification of uncertainties associated to equations of state parameters, 3) a Monte Carlo pro-cedure for the propagation of property uncertainties through the cycle process onto the model output uncertainty, and 4) novel strategies for the selection of working fluids under property uncertainties, in particular a new reverse engineering ap-proach based on sampling and uncertainty concepts. The framework is applied to different applications and case studies from industrial project partners.
Novel group contribution based property models are developed for the estima-tion of flammability-related properties (e.g. the lower flammability limit) of work-ing fluids. Compared to existing models, the ones presented here show a higher accuracy, are simpler to apply and provide every prediction value with its corre-sponding uncertainty range (with 95% confidence). The study also reveals that group contribution methods can suffer from parameter identifiability issues charac-terized by a significant correlation between estimated parameters. Hence, in order to ensure reliable estimation, reporting the 95% confidence interval of the model predictions is important.
In a second application it is shown how the uncertainty propagation of two types of equations of states, cubic and PC-SAFT, can be compared in the context of an industrial organic Rankine cycle, used for the recovery of waste heat from an engine of a marine container ship. The study illustrates that the model structure is vital for the uncertainties of equations of state and suggests that uncertainty becomes a criterion (along with e.g. goodness-of-fit or ease of use) for the selection of an equation of state for a specific application.
Furthermore, two studies on the identification of suitable working fluids for thermodynamic cycles are presented. The first one selects and assesses working fluid candidates for an organic Rankine cycle system to recover heat from a low-temperature heat source. The ranking of working fluids can be significantly differ-ent based whether the mean value or the uncertainties (e.g. the lower bound of the 95%-confidence interval) of the model output are considered. Hence, uncertainty analysis with respect to the input property uncertainties is a vital tool for model analysis and fluid selection.
In the second fluid selection study the novel reverse engineering approach based on sampling techniques and uncertainty analysis is applied to identify suitable working fluids for a industrial heat pump system, used to recover heat from spray-drying air in dairy industries. The novel reverse engineering approach provides a valid alternative to computationally demanding optimization approaches and al-lows to take into account property uncertainties.
The outcome of this thesis asserts that property uncertainties should be taken into account for process simulation applications, in order to support the model-based and reliable decisions on process fluids and process design.