Reverse Engineering of Working Fluid Selection for Industrial Heat Pump Based on Monte Carlo Sampling and Uncertainty Analysis

Abstract

This study presents a novel methodology for the identification of suitable pure component working fluids for heat pumps. Two challenges are addressed: the difficulties in solving a complex product-process design problem and making it accessible for practical applications, as well as the impact of the working fluid property uncertainties on the solution. A Monte Carlo sampling is applied to generate sets of different property parameter combinations (virtual fluids), which are subsequently evaluated in the heat pump process model. The distance between the property values of the virtual fluid and the uncertainty bound of the properties of real fluids (collected from a database) are calculated. The fluids that are closest to the top-performing virtual fluids are further analyzed through evaluation in the cycle and subsequent uncertainty propagation of the respective input property uncertainties to the model output uncertainties. The methodology has been applied to an industrial heat pump system used for waste heat recovery from a spray drying facility in the dairy industry. To remain focused on the validation of underlying concepts of the methodology, the study considered screening only among cyclic hydrocarbon working fluids. The compounds identified by the methodology had a low global warming potential (<0.1) but a high flammability (lower flammability limit <1.8 vol %). Cyclopentane showed the best performance with a coefficient of performance of 3.06 ± 0.05. The sampling-based reverse engineering method identified top performing working fluids, but avoided solving complex and computationally demanding molecular design problems, and took into account the real fluid property uncertainties.

A Multi‐Criteria Study of Optimal Working Fluids for High Temperature Heat Pumps

“…Comparing the results of the mapping step to findings of prior CoMT‐CAMD‐like studies is difficult: McLinden and co‐workers 4, 5, 34 study a much lower target temperature level for the provided heat of T cond = 40 °C and consequently consider only molecules with a critical temperature below 400 K. Similarly, Roskosch et al 35 consider a maximal value for T cond of 60 °C. Frutiger et al 36 consider very similar temperature levels as in the current study but focus only on the value for COP and only on cyclic hydrocarbons. Still, there is an important overlap in the lists of proposed molecules: cyclobutane is also ranked as one of the best performing molecules in their study.…”

“…CoMT‐CAMD‐like approaches have been applied to identify optimal working fluids and process specifications for heat pump systems in several studies 34–36. From the approach of this study, they differ in the physical property model that is used to determine thermodynamic properties of the working fluids, especially in the approach to model the value for , the number of parameters for the physical property model as well as for the process specifications that are included in the optimization, the objective function, the heat pump cycle configuration as well as in the specified boundary conditions such as the temperature levels at evaporation and condensation.…”

Heat pump working fluid selection—economic and thermodynamic comparison of criteria and boundary conditions

“…Different studies, e.g. (McLinden et al, 2017), (Frutiger et al, 2018) and (Cignitti, 2018), have performed comprehensive screenings for specific applications. McLinden et al (2017) summarizes the works on working fluid screenings among extensive databases and concludes the options for working fluid options to be limited.…”

“…As the estimation was expected to provide sufficient accuracy for most of the fluids, we recommend not to exclude mixtures with estimated interaction parameters on beforehand, but to conduct an uncertainty analysis as part of the screening evaluation in case such mixtures indicate promising performance. If experimental data becomes available, the mixture could be re-evaluated using updated interaction parameters or an uncertainty analysis as described by (Frutiger et al, 2018(Frutiger et al, , 2016) could be conducted.…”

Computer-aided molecular product-process design under property uncertainties – A Monte Carlo based optimization strategy

“…Each candidate compound can be ranked according to the uncertainty range of the process model output subject to the compounds property uncertainties [33]. Most recently Frutiger et al [36] introduced a novel reverse engineering approach for the fluid selection process based on Monte Carlo sampling: A Monte Carlo sampling algorithm has been used to generate sets of different property parameter combinations (virtual molecules), which are subsequently evaluated in a process model. The distance between the property values of the virtual molecules and the uncertainty bound of the properties of real compounds (collected from a database) are calculated.…”