The present design practice for dehydration tanks and vessels is typically based on a plug flow/Stokes' law approach, which is considered as unrealistic. Instead, these tanks may be considered as deep layer settlers, where the actual separation takes place in a dispersion band located between single-phase oil and water layers. The variation of the thickness of the dispersion band with the throughput of the settler is a characteristic function of the heaviness and stability of the feed, and depends on the operating conditions. In a new design philosophy developed by Shell Research these settling characteristics are being used to establish the operating window of primary oil/water separators. A 100 litre model settler has been built to study the settling characteristics of crude/water dispersions on a representative scale. Batch settling experiments have been carried out with crudes of various gravities to establish the relation between nominal settler capacity, crude viscosity and dispersion quality. This information was complemented with operating data of existing dehydration facilities to obtain a data-base for design. This paper discusses the proposed design method and the results of the model settler experiments. A comparison will be made with the conventional approach and the first applications will be illustrated. An Alternative for Stokes' Law Present design rules for crude/water separators are based on variations of Stokes' law, which relates the maximum allowable oil flow rate to the physical properties of the crude and an arbitrarily chosen cut-off droplet diameter d (usually 200), as follows: (1) where Qoil is the oil flow rate, A the horizontal cross-sectional area of the separator, vs the settling velocity of a water droplet of size d, p the density difference between water and crude, and the viscosity of the crude. Stokes' law would be valid for unhindered settling, which means that it is implicitly assumed that the dispersed phase concentration is low, say not more than 5-10%. During most of its life a dehydration tank/vessel will not see such dilute dispersions. When the dispersion becomes more concentrated the droplets will start to hinder each other, resulting in a lower settling velocity. On the other hand the droplets will start to grow by coalescence, which leads to faster settling. At present it is not yet possible to predict the resulting settling process on the basis of these droplet/droplet interactions. An alternative to such a micromechanical approach is to consider the dispersion as a continuum, and to describe its behaviour on a more macroscopic basis. This is a customary approach for the design of deep layer settlers in the process industry, firmly based on the extensive investigations of Barnea and Mizrahi and their predecessors. The actual separation in a tank/vessel takes place in a two- phase zone, the dispersion band, located between an oil and a water monophase, see Fig. 1. The dispersion band is bounded at the top by the settling front, which is determined by the smallest water droplets that can just settle against the rising oil flow, and at the bottom by the coalescing front where the largest water droplets merge with the water-phase. During stationary operation the dispersion band is in equilibrium: both fronts approach each other with a velocity that just matches the superficial velocity of the incoming fresh feed, when the throughput increases, the dispersion band expands to establish a new equilibrium at a higher separation capacity: expansion means less hinder for small droplets, and more residence time for coalescence. The resulting dynamics of the dispersion band, the variation of the thickness HD as a function of (gross) throughput per unit cross-sectional area Q/A, is characteristic for each feed, and can be described by the equation: (2) P. 669^

No abstract

Tests were carried out to investigate the potential of new Zeta Dynamics internals for the optimisation of the water handling capacity of the Draugen primary separators. To this end these new internals were installed in the Draugen test separator and an assessment was made of its separation performance before as well as after the installation. Since the production at Draugen is still practically dry, the advent of water production was simulated by flowing heated seawater through the separator. The tests were carried out on the wellstream of one of the horizontal platform wells, and water was injected upstream of the choke valve using the high mixing intensity in the valve to disperse the water in the crude. The tests consisted of varying the crude and water flow, and measuring the distribution of water and oil upstream, inside and downstream of the separator, via sampling and radiological techniques. It could be concluded that the modification of the separator internals has not led to an obvious breakthrough in separation performance. The new internals improve the flow distribution, as claimed by the vendor. This appears from the sharper interfaces in the tests after the modification, and from the reduced spread in the residence time distributions. However, the new inlet internal has a negative effect on the phase distribution of the incoming feed. This appears from the increasing mixing of gas, crude and water in the separator with increasing flowrate, this despite the simultaneously decreasing stability of the crude/water dispersion and the existing gas/liquid stratification at the inlet. It can be conjectured that the originally installed Shell 'Schoepentoeter' inlet gives less resistance to the flow and may even favour coalescence and preseparation due to centrifugal acceleration. The test results show that the size of the water droplets at the inlet is correlated with the water concentration and the pressure drop over the choke valve. However, the correlation is different from the one established from laboratory experiments, which is probably due to the different degree of premixing in both cases. The density profiles in the separator give evidence for the presence of a dispersion band, varying in size with process conditions. The predictions of separation efficiency as a function of flow rate and feed droplet size according to the SRTCA dispersion band expansion model (ref. SPE 38816) are in agreement with the test results. Finally the future dehydration capacity of the Draugen installation was estimated by combining the validated separation efficiency correlations with calculations of the droplet size expected at the prevailing operating conditions. P. 121