Towards the realization of subsea factories: Thermodynamics of petroleum fluids relevant to subsea processing

The aim of this study was to generate useful contributions for Equinor in their pursuit of commissioning fully-fledged underwater oil & gas processing facilities. Although subsea installations face many challenges, they are also able to leverage favourable processing conditions which are not practically achievable in traditional onshore and offshore facilities.

While Equinor are tackling a broad array of projects, we were focused on high-pressure subsea natural gas dehydration and specifically the study of the thermodynamic behaviour of relevant petroleum fluids. In this endeavour we have presented new work in the following fields: • Multicomponent two- and three-phase equilibrium data • Association schemes for the description of glycols within the SAFT framework • Uncertainty and sensitivity analyses for process simulations At our laboratories in Denmark, an experimental apparatus was modified for the quantification of C1-nC6/nC7-H2O phase distributions (VLLE) at 303-323 K.

From the process perspective, these data relate to the pre-separation step which occurs prior to dehydration. Due to the changes made to the apparatus and specifically the analytical equipment, it was necessary to validate the accuracy of the system. Good agreement was achieved with several binary VLE sources from the literature, for both vapour and liquid phase analysis.

The experimental uncertainty (in terms of composition) was estimated at ±9% using a 0.95 level of confidence. While no three-phase ternary sources are available for direct comparison with our newly measured data, comparable uncertainty ranges were observed for the quantification of all components in all phases for similar experimental studies.

Two-phase data directly relevant to the natural gas dehydration step were measured during an external research stay at Equinor in Norway. C1-MEG-H2O and (natural) gas-MEG-H2O systems were investigated for temperatures in the range 288-323K and pressures of 60 and 125 bar. The uncertainty was found to range from ±2% for major components up to ±42% for trace components in the natural gas.

In modelling of the ternary data, it was found that CPA provided superior prediction over the SRK-HV equation of state. The 3C, 4E and 4F association schemes were proposed for MEG and evaluated for the CPA equation of state. A new 4C parameter set was also proposed, and the importance of using raw experimental data in parameter regression was highlighted.

The bootstrap method was used quantification of the parameter uncertainty. Overall results showed that no scheme is universally superior, but improved prediction over the literature parameters was achieved in all cases. The 4F scheme performed best for ternary data prediction. Combined parameter uncertainty and process input sensitivity analyses were performed for simplified natural gas dehydration configurations.

Using Monte Carlo simulation, distributions were generated for process outputs such as product gas quality. Process conditions were optimized to achieve the desired specifications with a 99.7% confidence interval. Additionally, to the scientific outcomes of this work, there were also contributions made in a ‘co-supervisory’ role to the theses of two students (one master’s and one bachelor’s) during this PhD.

This work has already resulted in four publications, with an additional three manuscripts in preparation for submission.