Thermodynamic modelling and simulation of mercury distribution in natural gas

Abstract

Mercury and its species occur naturally in all fossil fuels, such as natural gas (NG), crude oil, and coal. The concentration of mercury in oil and gas varies depending on source, but it is usually of the order of a few parts per billion (ppb). However, even at these very low concentrations, mercury and its species can cause significant problems during oil & gas processing and, therefore, its levels in a plant must be monitored. Mercury is toxic to living organisms and, for this reason, strict regulations are in place regarding mercury emissions to the environment from industrial activities. In addition, mercury can cause catalyst poisoning and corrode the equipment through various mechanisms, such as Liquid Metal Embrittlement (LME)[1]. Indicative of the risk that Hg poses to process is the fact that until today about 10 industrial accidents have been recorded, which were caused by corrosion of equipment by mercury. For the proper management of mercury in oil & gas treatment plants, it is necessary to know the distribution of mercury in the different phases, e.g. gas, liquid, aqueous, during drilling and topside treatment processes. Scientific research in this field is currently active. Although some thermodynamic models describing the distribution of elemental mercury in NG have already been proposed, there are still aspects of the issue that are not sufficiently covered in the literature. For example, although in the vapor and liquid streams of the NG processing plants various mercury forms other than elemental, such as HgS, HgCl2, MeHg, Me2Hg etc. have been identified, their distribution has not been described so far with any thermodynamic model. The existence of other forms of mercury apart from the elemental indicates that it may also participate in reactions during NG processing, which have not been investigated in the open literature nor have they been included in any model. The development of such a model becomes even more complicated if one takes into account the adsorption/chemisorption of mercury on piping and equipment walls, and also the great variation in pressure and temperature along the natural gas value chain. The aim of the thesis is to develop a thermodynamic model that can accurately describe the simultaneous chemical & phase equilibria (CPE) of mercury in natural gas, and its application in the simulation of Hg distribution in natural gas processing plants. Towards this, the UMRPRU EoS/GE model is extended to mixtures of mercury with compressed gases (CO2, N2), hydrocarbons, water, and polar compounds that are often encountered during oil & gas processing, such as amines, glycols and alcohols. For comparison purposes, the widely used cubic EoS SRK and PR are also employed. To ensure that the models correctly predict the vapor pressure of pure mercury, different functions for their attractive term are examined. For UMR-PRU and PR the Mathias-Copeman a-function is proposed, while for SRK the afunction by Twu is employed. Pertinent a-function parameters are fitted to pure mercury experimental vapor pressure data with average absolute relative deviation (AARD) lower than 1%. Afterwards, model interaction parameters are fitted to experimental Hg solubility measurements. For the cubic EoS, generalized correlations for the binary interaction parameters are developed for hydrocarbons, while for polar compounds temperaturedependent BIPs are determined. The overall results show that UMR-PRU yields the best results in binary hydrocarbon and polar mixtures of mercury, while it also yields the lowest deviations in most multicomponent hydrocarbon mixtures and in all polar multicomponent mixtures. In order to study the possible reaction between mercury and hydrogen sulfide in natural gas (Hg0 + H2S  β-HgS + H2), the UMR-PRU model is also extended to mixtures of hydrogen with compressed gases (CO2, N2), hydrocarbons, water, and polar compounds. For comparison, the PPR78 model is also employed. The ability of PR to predict pure hydrogen properties is checked, and the Soave expression for the attractive term is found to yield the best results, while also ensuring that the a-function is consistent. UMR-PRU model interaction parameters are then determined by fitting binary vapor-liquid equilibrium data for hydrogen binary mixtures. It is found that UMR-PRU shows a lower overall deviation in bubble point pressure (8.1%) than PPR78 (13.2%). Both models are also employed for predictions in multicomponent hydrogen mixtures with hydrocarbons and compressed gases, with UMRPRU yielding the best results. After successful model extension to mercury and hydrogen mixtures, UMR-PRU is employed for calculating mercury saturation concentration in typical hydrocarbon fluids. For this purpose, the multiphase flash algorithm that was developed in this work is employed, which can handle systems that contain up to four phases: vapor-liquid hydrocarbon-aqueousmercury. The results show that mercury solubility in the various phases increases exponentially with temperature and generally increases in the order aqueous < vapor < liquid hydrocarbon phase. The effect of pressure on mercury solubility in the different phases is also examined, and results show a weak dependency in the liquid hydrocarbon and aqueous phases. On the other hand, Hg0 solubility in the vapor phase is found to decrease with pressure, until a plateau is reached. Phase composition is found to play an important role and different behaviors can be observed, e.g. in fluids involved in early-stage separation processes from those that can be found in the condensate stabilization train of a gas processing plant. The second point of focus in this work is the theoretical study of the reaction between elemental mercury and H2S in natural gas, which could provide an explanation for the origin of β-HgS solid particles found in condensate tank sediments. Chemistry dictates that mercury has a high affinity for sulfur and its compounds, and H2S is the most abundant sulfuric compound in natural gas, so a reaction between them is deemed reasonable. Both cases of vapor and liquid phase reaction are examined by calculating the pertinent equilibrium constants. Then, the simultaneous chemical & phase equilibria in the same fluids were solved by employing the Gibbs energy minimization algorithm developed in this work. The UMR-PRU model is subsequently employed for simulating mercury distribution in an existing offshore natural gas processing platform. For comparison the SRK-Twu model is also used, and model results are compared to field measurements regarding mercury concentration in selected streams. For the purposes of this study, a simplified version of the process is implemented in UniSim Design R460.2 and the distribution of mercury in the various streams is examined. The effect of the reaction between mercury and H2S is also studied. Different scenarios are considered, based on the presumed amount of mercury in the plant feeds according to mass balance calculations. Mercury partitioning in the TEG dehydration & regeneration process, as well as in MEG regeneration are also examined in separate simulations. The results in the case of no reaction show that both models yield very good predictions regarding mercury concentrations in process gases, but overpredict Hg levels in condensate fluids. UMR-PRU is found to yield the most accurate results for Hg distribution in aqueous streams, as well as in the processes involved in TEG dehydration & regeneration, and MEG regeneration. On the other hand, when the reaction is also included, it is found that the models yielded better results for Hg concentration in condensates, but deviate from the measurements in gas streams. In addition, UMR-PRU predicts an amount of produced solid β-HgS, which is closer to the expected value based on the field data. Considering the uncertainty of measurements concerning Hg concentration in liquid samples due to various experimental challenges, it is deemed that UMR-PRU yields the best overall results, while it is also capable of describing processes involving polar compounds, such as TEG dehydration & regeneration, where classical cubic EoS perform poorly. Finally, the developed CPE algorithm is applied for the study of complex mixtures involving non-reactive and reactive azeotropes. Such mixtures are commonly encountered in the chemical and petroleum industry, and require advanced thermodynamic tools that can accurately predict their equilibria. Such tools are important in order to determine the feasibility of separation processes, such as reactive distillation. In this work, the CPE algorithm is applied for studying the MTBE synthesis from methanol and isobutene, as well as the synthesis of isopropyl acetate via esterification of acetic acid with isopropanol. The algorithm is coupled with classical activity coefficient models, UNIQUAC and NRTL, as well as with UMRPRU. The results show that the CPE algorithm is very robust, and that thermodynamic models coupled with the algorithm can successfully describe the chemical & phase equilibria involved in these systems, providing important information about the feasibility of separation processes.