Preparing a thermodynamic framework for CO2 capture process
Abstract
Summary of thesis:
Climate change, with more than 95% certainty, see IPCC 2014, is happening because of anthropogenic CO2 emissions. The easiest and fastest way to cut emissions is CO2 capture and storage. Post combustion CO2 capture is very easy to implement as there is no need to do much modifications to the existing power plant. Large scale implementation drags out mainly because of capital and operational costs, the latter mainly due to energy requirements. In order to decrease the energy requirements both the solvent and process can be optimized. In this development, having good thermodynamic models to predict the equilibrium, thermal and physical behavior of the system is an essential part.
Commercial softwares already exist that could be used for simulation of various processes. HYSYS, Aspen, ProTreat, ProMax, Pro/II and OLI electrolyte are among the most well-known and powerful simulation packages. In this thesis, in a comparative study, the performance of a selection of these softwares is compared when predicting VLE, density and viscosity of CO2-loaded amine solutions. Among them, the OLI electrolyte was found to have the best performance.
Existing software have limitations with regard to the selection of solvents and for new solvents it is necessary to have a thermodynamic framework available for rigorous modeling. An activity based approach was chosen and the UNIQUAC framework was selected based on ease of implementation and availability of some of the required interaction parameters. The solution of the reaction equilibrium in the liquid phase is done using the RAND method and a Gibbs energy minimization. Activities in the liquid phase are calculated using the UNIQUAC and Debye-Hückel models for taking into account both long range and short range forces.
Optimization of the model is done using a combination of two methods; a Pattern Search and a bounded simplex Nelder-Mead. The objective function for optimization was selected based on checking the result of using different ways to define the error value and aiming at preventing the model from both under-prediction and over-prediction.
Benchmarking of the model is done using CO2 absorption into the aqueous MEA system. Further the model was developed for AMP, Piperazine and for mixtures of the two. For AMP there were 2 problems: there were almost no published data for the AMP carbamate formation constant and no data for high temperature VLE of the system. For addressing the first challenge a computational chemistry model was used to calculate the temperature dependency of the carbamate formation reaction so only the absolute value at one temperature had to be optimized. Considering the second problem, heat of reaction measurements were used for tuning of the model using 2 different approaches for calculating the heat of reaction: the Gibbs-Helmholtz equation and the extent of reaction methods.
Piperazine (PZ) suffers from many of the same problems as the AMP system: lack of data for high temperatures and lacking measurements of several reaction constants, both absolute value and temperature dependency. The same approach as for AMP, based on using heat of reaction and computational chemistry for calculating the temperature dependencies of the reaction constants which were not available.
For the AMP-PZ system the biggest problem is the lack of reliable data and significant scatter in data from different sources.
For predicting molar volume and viscosity of the mixtures, a new approach based on using UNIQUAC and its derivatives is proposed, and is shown to have good capability in predicting defined excess values and the mixture values consistently.