## Modelling of Mercury (Hg) Distribution in a Natural Gas Processing Plant

##### Abstract

SummaryMercury (Hg) is a toxic chemical element, commonly known as quicksilver and hydrargyrum. Found in all hydrocarbon mixtures, mercury has been recorded in natural gas and oil fields all over the world. Generally significantly high levels of mercury concentration are found in the natural gas with light hydrogen isotopes of methane and at lacustrine dispositional environment. Mercury can cause significant problems to the processes of the gas industry such as aluminum heat exchanger failures, poisoned catalysts, contaminated product streams and gas leaks.Given the problems that Hg can cause to the industry and the human health, an investigation on the matter is required. The scope of this thesis is the development of a thermodynamic model that is able to predict the elemental Hg distribution throughout a natural gas processing plant. This way the most efficient way of dealing with this component and the problems caused by it can be detected and solved. To this purpose, the Peng-Robinson (PR) and the Soave-Redlich-Kwong (SRK) cubic equations of state (EOS) were examined by employing different expressions for the alpha parameter of the EOS. For this parameter, Mathias-Copeman and Twu-Coon pure components parameters were developed for Hg for PR and SRK EOS respectively by fitting them to vapor pressure data of elemental Hg. Except for those parameters, binary interaction parameters for Hg and hydracarbons (HC) were developed as well by fitting them to solubility data of Hg in binary mixtures with various HC. Three different types for those parameters have been developed, constant ones, temperature dependent ones, and also a predictive correlation to estimate them for binary mixtures of Hg with HC for which no data are available in the literature. The models have been also tested in a case study -a part of the Kårstø plant-. In total three different models have been developed and tested in this master thesis. One of them is based on Peng-Robinson and the other two on Soave-Redlich-Kwong cubic equation of state. The tests have three objectives. The first one is to see the effect while using different thermodynamic models on the prediction of the distribution and mole fractions of Hg throughout the process without the binary interaction parameters. The second one refers to the effect of the binary interaction parameters, once taken into consideration on the distribution and mole fractions of Hg throughout the process. The last one is to find the optimum type of binary interaction parameters that should be used for process simulations. It is mentioned and strongly underlined the fact that there is no possibility of verifying the results of the case study due to the lack of any field data.The results of the simulations can be summarized as follows:1. All models with no binary interaction parameters provide different results in terms of quantity regarding both the distribution and the mole fractions of Hg throughout the process. At the demethanizer the PR-MC model predicts that the top product of the column has two times the amount of Hg that the two SRK-Twu models suggest. The main difference however is that the PR-MC model predicts that there is more Hg in the top product of the depropanizer compared to the other two models, and also that after that distillation column almost the whole amount of Hg goes to the top products of the debutanizer and the Splitter. The two SRK-Twu models predict the exact opposite fate for Hg for the last two distillation columns and are in agreement with each other. Given the differences in the percentage distribution throughout the process, the same differences concerning the mole fractions of Hg appear as well at the same parts of the process.2. All models with binary interaction parameters -regardless of their type- provide similar results regarding both the percentage distribution and the mole fractions of Hg throughout the process. Therefore the constant kij parameters are proposed because of their advantage to assist in the stability of the algorithms that solve the simulations.3. All models agree that almost the whole initial amount of Hg that entered the process is removed by the time the debutanizer is reached by the flow. This is an important difference compared to what the results of the case with the kij parameters set as zero suggests. In addition all models suggest that whatever Hg reaches the debutanizer and the Splitter columns ends up in the top product streams of the columns.4. The mole fractions of Hg predicted by the models with the binary interaction parameters are much lower than what the models without these parameters predict. This however is to be anticipated up to a point because of the fact that there are important differences in the distribution of Hg throughout the process.5. It turned out that there was no real difference between each type of kij parameters. Therefore the constant ones are proposed, because they assist in the stability of the algorithms solving the simulations.