chilled mirror or software : this is...

[OldDoc]

we have a chilled mirror system for monitoring natural gas (mainly C1) dew point, it should provide two/three degrees (Celsius) of precision, we have also a in-line gas chromatograph for composition (C1-C6+) , I got a suggestion to put the composition on a computer, calculate the dew point and remove the old instrument but I am not sure about the final precision of the alternative, someone has compared the two systems ? I see that dew points calculated with computers are sensible to very small fractions (ppm) of heavy hydrocarbons and that could increase considerably the dew point,
thank you.

 

[dcasto]

the Gas Processors association

http://www.gpaglobal.org

has papers on this topic. Go to publications and annual meeting papers.

Generally, the mirror is the standard that you must calibrate your GC and it's EOS to. The issue in using a GC and EOS is charcterization of what C6+ is. The issue with the mirror is the training of people that can use it correctly.

 

[PaoloPemi]

I agree with dcasto, the mirror is usually the standard and 1-2 degrees is generally considered a good accuracy for dew point. A GC/EOS system offers the advantage of many different properties (VLE, critical, transport) calculated from composition, the EOS predictions are very accurate but depends from composition and a GC stops usually at C6, also the EOS procedure can be very sensible to small fractions in C6+ where the mirror tends to average values, however with EOS you can tune the sensibility by selecting a line of liquid (or vapor) fraction different from 0 or adjsut parameters to fit some data, in that case the results are comparable, to evaluate the influence of heavy components on dew points (and cricondentherms etc.) see for example

http://www.prode.com/en/dewbubblenaturalgas.htm

there are many papers on this topic and also results of comparative analysis which may be of interest for you.

 

[OldDoc]

thank you for the comments,
we have a in-line mirror which is expensive and difficult to maintain, also I am not sure about reliability (laboratory gives quite different results, 3-5 degrees difference), we are considering alternatives, our in-line GC is limited to C6 but we get regularly C6+ from laboratory. Clients would prefer a in-line mirror as reference but I see that GC + software solution has gained wide recognition and acceptance and I am looking for information supporting the alternative.

 

[dcasto]

there is nothing to support it. As I pointed out, it's being studied by the GPA. You could negotiate with the pipeline company and come to an agreement. I had one pipeline that used a cheapie EOS software and the charcterized the C6+ in their own way. We were delivering gas at 45 Degrees F. Their GC and EOS said the gas had a dew point of 61 degrees. I pointed out that their EFM was reading 45 F, so by definition, the gas going throught the GC had a 45 F dew point. They let us flow.

 

[OldDoc]

thanks.
I must agree with your comment, there are no ways to get the real dew point as the mirror needs some content of condensate (70 mg.m-3 is a std. value) to react, also mirrors can show errors of 5-7 C (MCMI v ACMI see the NPL REPORT AS 3), GC-EOS packages in general could be very accurate but they suffer from poor composition characterization, when tuned they need a mirror as reference, however this could be the laboratory's intrument.
In our case the GC manufacturer is proposing a quite expensive EOS package (>10000 US) so there are not real economical reasons to support this EOS solution... instead We are considering to install a EOS software within DCS and fix inputs for C6+ with the result from laboratory analysis, the preliminary results are promising, we plan to evaluate this solution for one / two months, collect results and create a quite large base of data to discuss with the clients, that could be a good starting point...

A think to mention : the EOS package would offer the advantage to calculate in realtime many properties that at present we obtain from our process simulator, I am not sure about final accuracy (must test and compare different software solutions) but this could be very useful for us...

A final note, we have evaluating different software packages and comparing dew point / cricondentherms as calculated with Soave Redlich Kwong and Peng Robinson, results are quite similar (we use our process simulator as reference) despite very large differences in price, this really surpises me.

 

[MortenA]

It is a very "politica" subject

The way i see the problme is that:

Old emperical data is used to convert measured dew point to a water content e.g. provided by GPSA. These data were produced in '50-'60

When you move back and forth using this methode it "adds up".

However: The European Gas research institure (GTI) has developmed a new model called GERG based on a modified SRK and using very precise data produced in the '90:

http://www.gastechnology.org/webroot/app/xn/xd.aspx?it=enweb&xd=gtihome.xml

BUT - this will predict a much higher water dew point if you take your "old" max water content. In reality the imo the dewpoint remain the same but the water content has been much lower than prediced when using the old models. The GERG model has higher dewpoint at higher pressure than predicted by the "old" models.

BTW: SRK and GSPA etc yields approx the same results - most likely because they are based on the same experimental data.

Best regards

Morten

 

[PaoloPemi]

if I remember correctly the old GERG 88 was a modification of Peng Robinson (with Kij = 0.5 for water) but the new GERG 2004 is a very different beast, indeed it's very accurate for gas densities (verified to be more accurate, for test mixtures, than AGA 8) and can work liquid phase (AGA 8 cannot) so (within its range of application) has been proposed for VLE of natural gas mixtures, tests (against mirror based instruments) are encouraging but we are speaking of errors of 1-2 degrees for Peng-Robinson and Soave Redlich Kwong (with the same composition as input), well below the errors induced by C6+ characterization.... so while I would suggest AGA 8 or GERG 2004 for calculating densities of natural gas mixtures I think that gas dew point is more on a political discussion side...

 

[dcasto]

PaoloPemi, that is not true. Here is a table to review

At 600 psia and 92.5% C1, 5% C2, 1 % C#, .5% iC4, .5% C4, and a .5% C6+ we get these dew points.

srk pr
C6+ = C6 46.7 45.9
C6+ = 50% C6, 50% C7 71.3 72.9
C6+ = GPA's recommended 60%, 30%, 10% 77.3 78.5

there is a 30 degrees defference from these 3 methods and a 5 degree difference from two proposed characterisation methods, more than enough to shut in a delivery point. Hence the political basis for arguments and why the chilled mirror is still the standard.

 

[MortenA]

dcasto

It seems to me that the big difference is not related to PR vs SRK - but more to how the "tail" is defined?

(btw for GPA what is the 10% - C8?)

Best regards

Morten

 

[dcasto]

yeah, the EOS isn't as important as the definition of C6+, but if you are using PengRobinson and the pipeline company is using RK and they are shutting in 100MMSCFD of gas, does it matter, YES!

The GPA "recommends" that C6+ be described as 60% hexane, 30% Heptane and 10% Octane. This follows the typical gas analysis where each subsequently larger molecule has 1/2 the previous molecules mole %.

 

[PaoloPemi]

dcasto,
I mentioned "errors of 1-2 degrees for Peng-Robinson and Soave Redlich Kwong (with the same composition as input)", this has been verified for usual natural gas mixtures and I believe it's out of discussion; characterization method for C6+ (when full composition analysis is not available as with GC instruments) is not easy to define as different methods can generate different results, for example fitting C6+ content to one dew point temperature can simply give wrong results at different pressures... perhaps one can fix the C6+ components (as proposed by OldDoc) but I fear that there is not a safe way to follow.

 

[dcasto]

PaoloPemi, If you take the PR versus RK off the table, then what will you do if I shut you pipeline down for non conformance? It's still an issue. The contracts must state what EOS and what characterization will be used by both parties.

Southwest Research Institute in Texas, US is working on another mechanical method based on the chilled mirror, but with lasers reading the dew point. This system could solve it all.

 

[PaoloPemi]

the point (on which I think we agree) is that a EOS (PR,SRK etc.) to give results equivalent or better than a mirror needs the full composition, the way one estimates C6+ (usually not svailable with a in-line GC) can introduce very large errors, the different methods used to characterize C6+ should be specified by contract differently there is room for long discussions where everyone has some reasons, EOS is just a mathematical tool and results depend from inputs, data fitting works only when input variables are reasonably specified, differently you get wrong results, all the same...

 

[dcasto]

yes we agree. My belief is that it all comes down to contracts or agreements of the industry. In the US there was a big issue over imported LNG and dew point. The regulatory agency did not intervene and just said to all parties, "work it out, its not our issue".

 

[maddocks]

We're constantly arguing over the chilled mirror (Chandler) results because it is a subjective test open to interpretation by skilled and non-skilled workers. I have look at zillions of chilled mirrors with operations teams and we seldom agree on the dewpoint. Now, maybe we disagree because we want to find the answers that support our case.

We have been working with Bartec Benke - a german firm that claims to have an electronic version of the chilled mirror concept. Might be quite a bit better, but it has not yet been completely accepted by pipeline companies.

For the record, I usually look at the mole weight of the C7+ sample and then adjust the fraction of c7 and c8 until the mole weight matches up.

 

[PaoloPemi]

I have no experience with the mirror devices you mentioned but I used several others and I can easily agree that the main (and perhaps only) advantage is the direct reading (based on a physical principle) vs. the calculated lines of constant liquid (or vapor) fraction by a software package based on composition, I think the availability of a series of lines at, say, 10, 100, 1000 ppm of condensate much more illustrative of the process so I prefer to adopt the software approach (as mentioned in a different post there are tools which offer -in my opinion- excellent features compared to cost) however there is the problem that you need the full composition to feed the software and that is usually not available from in-line GC, an agreement with the client is usually required.

 

[OldDoc]

the cost of the software is not a problem providing it can replace the mirror which has a quite large cost due to maintenance...
We have tested two software packages, once properly tuned (settimg the same parameters) they give exactly (< 0.5 K) the same values (for dew points, cricondentherm etc.) as our simulator, no convergence problems.
Now we are attempting to correlate C6+ (on the basis of laboratory analysis) then we hope to have elements to persuade the client...

 

[maddocks]

OldDoc - can you keep us informed of how this works out? Sounds like you may be on to something.

 

[dcasto]

I've looked at using WinSim's dll link to their engine. You could take real time SCADA GC data through their flash and calculate dew points. You can also link GC outputs through it and get dew points. It's never been a real issue so, I've not fully explored this.