May-2013

Gas treating simulation — a holistic perspective part 3: co2 removal in an LNG plant

This is the final article in a three-part series dealing with CO2 removal in a variety of real-world cases in which the behaviour of the gas treating plant is somewhat counterintuitive.

Nathan A Hatcher, R Scott Alvis and Ralph H Weiland
Optimized Gas Treating

Article Summary

Part 1 focused on a treating problem experienced in the CO2 removal section of an ammonia plant. There, a seemingly-small departure from an operating condition recommended by the process technology’s supplier resulted in failure to treat adequately by a huge margin. In Part 2, we looked at the surprising relationship between reboiler duty and solvent circulation rate in a pilot plant for CO2 capture operating to meet a specified fractional removal. To understand the behaviour in this case, one had to develop and appreciate the difference in operating philosophy between conventional deep CO2 removal and the removal of a limited amount of CO2, specifically only 90%. In this Part 3, we analyse the removal of CO2 with a piperazine-promoted MDEA-based solvent in an LNG plant where the target gas purity is <50 ppmv CO2.

As will be seen once again, developing a good understanding of the behaviour and performance of the treating plant depends on being able to look at either the whole plant or a specific tower and not focusing too soon on some specific detail or set of details. Although especially true in troubleshooting, this philosophy applies equally well to grass roots plant design, the focus of this article. The primary tool for the discussion in this article is mass transfer rate simulation, specifically the ProTreat simulator, which creates a virtual plant on a computer. The virtual plant is a mirror image of the real plant on a scale and to a level of detail that permits examination of detailed performance and behaviour. As will be seen, what such a model is capable of revealing can be quite intricate and complete, and very satisfying from an engineering science perspective. Perhaps of more practical importance, it can lead to the solution of a difficult troubleshooting exercise, on the one hand, or to the selection of correct operating conditions for a new plant or column on the other.

CO2 removal using MDEA promoted with piperazine
although MDEA is alkaline and therefore increases the reactive OH– ion concentration, MDEA itself does not react with CO2, so MDEA has very limited ability to influence the CO2 absorption rate. At best, MDEA might be said to catalyse CO2 hydrolysis by providing a more alkaline environment than water, but its inherent lack of reactivity makes it incapable of removing CO2 quickly. For this reason, MDEA simply cannot reduce CO2 to low concentrations in reasonable packed column heights or numbers of trays. This makes it an excellent choice for slipping CO2 but a terrible choice even for moderately deep CO2 removal. Piperazine, on the other hand, reacts extremely rapidly with CO2 (some 10 times faster even than MEA) which makes it an excellent promoter when used in relatively small concentrations with MDEA. Thus, piperazine-promoted MDEA is quite commonly used in LNG applications because it allows deep removal but has the low energy benefits of MDEA. It is offered by major solvent vendors under a variety of trade names.

LNG absorber
This case is a sensitivity analysis of the design of an absorber in an LNG plant to treat a feed gas containing mostly methane (84%), with 10% ethane, 4% propane and 2% CO2 on a dry basis to a specification of <50ppmv CO2. The proposed absorber contains 60 feet of IMTP-50 random packing and it was sized in each simulation for 80% of flood. Three sets of simulations were run at a series of solvent rates, but with each set having a constant value of CO2 lean loading, as shown by the legend in Figure 1.

At each lean loading, the absorber fails to treat adequately if the solvent rate is too low. This is as one should expect, because the solvent has inadequate capacity when the flowrate is low. This makes the column rich-end pinched and allows a significant amount of CO2 to pass through the column without being absorbed. The temperature profile at 500 gpm is shown in the left-most plot of Figure 2. Increased solvent flow improves treating, and if the lean loading is low enough there is adequate absorbing capacity to achieve <50 ppmv CO2. But as solvent flow is increased further, the temperature profile deforms into the shape shown in the middle plot of Figure 2 for 800 gpm. At this flow rate, the absorber is bulge pinched in that the temperature in the central part of the column is so high, only the ends are effective in removing CO2. The centre part of the column does nothing, and the column behaves as though it has perhaps 30 or 40 feet of packing, not the 60 feet that is really there. As the solvent rate is increased further, the temperature bulge gets pushed further down the column, and the absorber becomes lean-end pinched, where treating is determined primarily by the solvent’s lean loading, i.e., by vapor-liquid equilibrium conditions at the lean (top) end.

In this particular case, the treated gas was 40–45 ppmv CO2 over the flow range from 600 to 700 gpm but the way to respond to the gas going off-specification may not be to increase solvent rate, but rather it may be to decrease it, or perhaps to increase reboiler steam or hot oil flow. Without a detailed operating diagram such as the one in Figure 1, operations could probably not do much more than just guess at the correct response and hope for the best. And if the engineering contractor uses an equilibrium stage simulator of whatever ilk (ideal stages with Murphree efficiencies, or ideal stages with kinetic corrections and user-estimates of ideal stage residence times and stage thermal efficiencies), none of this would be apparent at all, and the design would have an unwelcome element of uncertainty and surprise.

The behaviour of trayed columns is a little different from ones containing packing. With packing, as the solvent flow is increased, the wetted, interfacial area rises as well, and the mass rate therefore increases with solvent flow for this reason. With trays, gas-liquid interfacial area for mass transfer is only a relatively weak function of liquid rate and the performance curve typically looks like Figure 3. (Note the logarithmic scale.) There is no maximum because the liquid-rate dependence of area is insufficient to drive higher absorption rates and torment the central, flat region into becoming a peak. Nevertheless, there are still lean-end, bulge, and rich-end pinch conditions at the low, medium and high solvent rates like those shown in Figure 2. Under the conditions for the simulation results shown in Figure 3, none of this really matters; however, if the lean loading were to become too high, one might have to operate at nearly twice the flow of very lean solvent to achieve the <50 ppmv specification.

The difference between packing and trays and, indeed the effect of packing type and size can be very important in the design of an LNG facility. If the simulator is not genuinely mass transfer rate-based, all of this will be missed. Trays and packing will all be treated as ideal stages; the differences between them will not be apparent (indeed, one packing will be as good as another and all packing will give the same results), the design will be subject to great uncertainty, and the plant may not work at all.