Most appropriate fluid/thermo package for TEG/MEG/Water systems 2

[ColourfulFigsnDiags]

Just wondering what the most appropriate fluid package for a MEG / water system. We are using HYSYS to model the system, and are not particullarly confident that the PR package we are using for the rest of the gas plant is applicable to the rich MEG streams.

Does anyone have any experience with modelling such systems?

Read the Eng-Tips Site Policies at FAQ731-376

 

[sshep]

You must use an electrolyte system (Aspen equivalent is ELECNRTL, but uncertain if Hysis has such an option). Now that Aspen has acquired Hysis, you have access to a good telephone support system to help you sort this out. This has been one of the few good things about Aspen Tech Inc.

best wishes,
sshep

 

[Zoobie]

I don't know how MEG is handled in Hysys but the new versions of Hysys have a Glycol Package which is supposed to work well for TEG/Water systems. I'm guessing that it probably would work as well for MEG but you should check with AspenTech support.

 

[ColourfulFigsnDiags]

Thanks all, I've logged a support ticket. My major concern is that using PR, when you mix a 50:50 steam of 50C ethylene Glycol with a stream of 50C water- the resultant mix temperature is 45C. But using an activity model like UNIQUAC, there is no temperature change at all. Is PR taking into account a "heat of dissolution" effect or is it simply a quirk of the EOS model?

Read the Eng-Tips Site Policies at FAQ731-376

 

[UmeshMathur]

ColourfulFigsnDiags:

The PR (Peng-Robinson) option is an equation of state that is valid for hydrocarbons, light gases, and petroleum mixtures. In general, an equation of state uses the same expression to calculate component fugacity coefficients in the vapor and liquid phases. In such cases, the liquid phase is assumed not to exhibit strong non-ideality in the sense of Raoult's law.

PR is certainly NOT recommended for glycol systems, as these contain highly polar molecular functional groups. Such mixtures show marked deviations from ideality. The activity coefficients are therefore far from unity and it is obligatory to use an activity coefficient model for the liquid phase. Examples are Wilson, NRTL, and UNIQUAC.

There are many published vapor-liquid equilibrium data for glycol systems (e.g., the DECHEMA data books, available in most good chemical engineering libraries). These can be regressed against the chosen activity coefficient model to develop the required binary interaction parameters.

Modern simulators already provide regressed interaction parameters for an activity coefficient model that you select. These are developed using the built-in physical properties for the pure components (most importantly, the vapor pressure), and thus you can use them directly. Otherwise, you would have no choice but to regress these parameters yourself.

My understanding is that Hysys does provide regressed binary interaction parameters for glycol systems (for Wilson, NRTL, or UNIQUAC). As zoobie has noted, the "Glycol Package" may already have selected the activity coefficient model and the associated interaction parameters for you.

I would pay absolutely no attention to any simulation results for these systems using PR, as the results would be pretty close to garbage.

 

[ColourfulFigsnDiags]

Thanks all, (partic Umesh!)
Well doing a sensitivity analysis revealed the following results.

Results, both inlet streams at 15.5C, 1 atm

[tt]
Results presented as: Mix Density (kg/m3), Mix Viscosity(cP), Mix Spec Heat(Btu/lbF), Mix Temp(C).

Mixing 1000kg/h Ethylene Glycol and 1000kg/h Water (50%wt EG)
Uniquac: 1076.1, 3.604, 0.8067, 15.5
Png Rob: 1079.4, 4.082, 0.8013, 11.2
Perrys : 1068 , 4.1 , 0.78 , N/A

Mixing 100kg/h Ethylene Glycol and 900kg/h Water (10%wt EG)
Uniquac: 1027.2, 1.361, 0.9621, 15.5
Png Rob: 1028.2, 1.400, 0.9867, 14.2
Perrys : 1012 , 1.4 , 0.97 , N/A

Mixing 900kg/h Ethylene Glycol and 100kg/h Water (90%wt EG)
Uniquac: 1111.9, 13.19, 0.6512, 15.5
Png Rob: 1113.4, 13.97, 0.6078, 13.5
Perrys : 1110 , 18 , 0.61 , N/A

[/tt]

In summary, both PR and UNIQUAC give quite similar results for everything other than the mixture temperature. This is concerning as you have all indicated that PR is not to be trusted, yet there is evidence that there is an endothermic heat of mixing reaction that occurs in EG /Water systems (ref 1 - thanks to DS of AspenTech), and the activity models will NOT take this into account.

Therefore I think I might need to include a heat of mixing into a tuned activity model to fully / accurately model this system.

ref 1: Matsumoto, Y ; Touhara, H; Nakahishi, K; Watanabe, N.
J. Chem. Thermodyn., 1977, 9, 801.

Read the Eng-Tips Site Policies at FAQ731-376

 

[UmeshMathur]

Hi, ColourfulFigsnDiags:

A few comments:
(1) It is hard to compute liquid density incorrectly far below the critical point, so that is not a good test of any model for liquid phase non-ideality.
(2) Viscosity is a transport property that has nothing to do with either an EOS or an activity coefficient model, hence this is also a non-issue.
(3) Specific heat is computed differently by an EOS as compared to what is normally done in an activity coefficient option. Again, this is not too dependent on the option chosen for the VLE thermo at temperatures far below the critical.
(4) Heat of mixing, on the other hand, can be predicted by an activity coefficient model. In fact, Hysys and Aspen Plus both have the ability to regress interaction parameters for such models using heat of mixing data. However, the dilemma that then arises is that predicting phase equilibria using interaction parameters obtained this way can be quite error prone. This is simply because these models are not sufficiently accurate for simultaneous prediction of VLE, LLE and heats of mixing. In specifying your option for calculating liquid enthalpies, you need to be careful to specify use of a heat of mixing approach.
(5) If accurate prediction of both phase equilibria and heats of mixing is necessary, then you ought to regress the interaction parameters simultaneously using both types of data. Such a procedure will help reduce the severity of the mismatch noted under (4) above, but neither VLE nor heats of mixing will likely be predicted with any great reliability. Unfortunately, there are few systems for which reliable data of both types have been measured, although the glycol-water systems are a notable exception.

There is a recent, interesting, and related thread that you may wish to refer to for a detailed discussion of how to fit heats of mixing with the UNIQUAC model: thread798-143708.

 

[ColourfulFigsnDiags]

Fantastic information Umesh! If you ever make it down to Melbourne, I'll have to shout you a beer or 3! Thank you!

Read the Eng-Tips Site Policies at FAQ731-376

 

[UmeshMathur]

Hi, ColourfulFigsnDiags:

There is a good discussion of this subject in Prausnitz, et al: "Molecular Thermodynamics of Fluid Phase Equilibria" (3rd Edition, Prentice-Hall, 1999) and also in Fredenslund, Gmehling, and Rasmussen: "Vapor Liquid Equilibria Using UNIFAC" (Elsevier, 1977). I learned a long time ago that heats of mixing are usually not sufficiently serious to hurt process simulation calculation accuracy too badly, especially distillation, for the vast majority of organic compounds. It's the VLE that really matters.

In VLE for glycol systems (e.g., dehydration applications), there is a preponderance of light hydrocarbons that are present way beyond their critical temperature, so the activity coefficient approach is not useful. To overcome the known deficiencies of a traditional cubic equation of state approach for highly polar systems, extensions were suggested (e.g., for the Peng-Robinson EOS) by Wong and Sandler. This effort has been ongoing for over 15 years. A good book on the subject is: Orbey and Sandler, "Modeling Vapor-Liquid Equilibria" (Cambridge, 1998). These extensions, with additional modifications, have found their way into most good process simulators, but it is still necessary that binary interaction parameters be fitted against VLE data for hydrocarbon-water, hydrocarbon-glycol, and glycol-water systems. One can then apply this approach with quite surprising success in glycol dehydration.

Thanks much for your kind offer. I have some buddies of almost 40 years standing in Sydney, and I have been thinking of taking the wife down under for a cool one to commemorate our 30th anniversary this summer, when temperatures in Houston can hit 35 C at 95% humidity!

 

[augiedog]

What are you trying to model? Typically there are two main uses for glycol in a gas plant:
a) If its MEG injection for hydrate control then the mass of gas will rule for the chiller duty required and the glycol temperature variation will have a minor effect on the overall refrig compressor horse power. The Hammerschmidt equation gives you the min. glycol flow and withdrawal concentration required.
b) If its for a heat transfer system using 50/50 MEG & water then its a closed cycle and so mixing doesn't enter into it and you can use the heat transfer medium vendor's data for heat & material balance in an excel spread sheet.

 

[ColourfulFigsnDiags]

Trying to model Rich-MEG to Lean-MEG regeneration system, storage and injection ... It turned out in the end that the variations between the models were not so important, but the knowledge to be learnt about the models themselves is great!

Oh and we will have to wait for that beer Umesh- I'm off to Scandinavia for about a year (out of the fryingpan into the ice!). Melbourne gets to 40+ a few times over summer, but fortunately it is only 20-30% RH (I think we peaked at 43.5 last year) ... Sydney, however, sounds like Houston!

Read the Eng-Tips Site Policies at FAQ731-376

 

[UmeshMathur]

ColourfulFigsnDiags:

Several process simulators have pre-configured thermodynamic "packages" for amine and glycol systems, e.g., ProMax (Bryan Research and Engineering) or VMGSim. Most designers use these, as they have been developed with the appropriate care to ensure accuracy and reliability.