MFA with substance layer and many periods

Hi again!
I have been busy again working with STAN and my question now is:
Is it possible at all to calculate a model in STAN which has several periods AND several layers (1 Good and 2 substance layers)? if the answer is yes, let me explain my current problem.

Now I have been working on a previous model, which was properly working in goods and all substance layers. I decided to continue my research with monthly steps instead of 1-year period to account for the different rain amounts during the year.

I changed the existent working model by making the 1-year period a 1-month period, I adapted the flows to match monthly quantities and then I added 11 more periods to make it a year. I filled the MFA data table with the correct information for goods and substances for all periods and I added the required relations for the substance layers.

Nevertheless, the calculation does not work properly. While calculating (I used all three available calculation modules) it takes very long time and at the end an "Error in calcualtion module IMPRSstanvienna12-7-231.exe" appears in the Trace Output panel.

The good news are:
- If I deselect the substance layers and only calculate the materials layer, it works great for all periods (even when at the beginning of the calculation an error message appers, as "No equation will be added for the relation (all Relations at substance layers)" but this does not affect the calculation in the materials layer).
- If I split the model in separate files for each month, the calculation works perfect as well for all layers, but then I have 12 files to work on.

The problem has obviusly something to do with the substance layers when all periods are in the same file. Have you had this problem before? I uploaded a file with five periods in case you want to take a look.

Thanks in advance for your comments.

Gabriela

Hi Gabriella,

You get results for your model if you change the settings of Kelly 2011 to:

• maximum number of non-zeros -> 1,000,000
• multiplier for maximum number of non-zeros -> 1,000
• epsilon -> 1E-9

There seems to be a structural problem in the combination of the system of equations that STAN generates internally and your input data. It is also very likely that this is the reason why Cencic 2012 cannot cope with the problem. I will have a closer look at the problem as soon as I have more time.

Oliver

Hi!
thank you so much for the answer. The changes did help and for the file with 5 periods (January-May) the calculation was successful. However, a funny thing happens when the number of periods increases: up to 8 periods are calculated sucessful for my model and then I get a simmilar message than before ("Error in calcualtion module IMPRSstanvienna12-7-231.exe") plus a message about the storage medium (see attached picture)


I did several tests, splitting the file with 12 periods in different ways to see if the problem was one period in specific, and it works fine with all kinds of combinations (Periods 1-8, 3-10, etc), as long as 8 periods were not exceeded. No matter which periods I choose, the calculation would only work for a maximum of 8 periods.

In any case my problem was almost solved because now I am working with only two files with 5 and 7 periods each, instead of 12 files with one period each! that easens up my work enormly.

In case you want to dig into the problem, I will upload my 12 period file.

Thanks again!

Gabriela

Hello,

a few months ago, I had my models working properly in STAN 2.5.1072 (after some adjustments to the calculation settings using Kelly 2011 as you can read in the thread of this topic)

Now, no matter what I do, they do not work anymore and I get a message of "smoothability state". I updated windows, also the Redistributable Library Package and I have MS Visual C++ 2010 Redistributable Package installed before.

I did not change anything in my files and the calculations settings are the same also. I just open them and they are still showing the calculation results of last time in October 2013, but when I try to recalculate them they dont work.

The new STAN version doest not work for me, because I have too many variables and equations in my model. But I really need to make some last flow corrections to finish the simulations for my PhD research (4 years work now).

My questions are:
Is there a way to make STAN 2.5.1072 work again?
or is the only solution for me to split my model into parts to be under the 2000 restraint limit of the new STAN?

I know I could also buy the unlimited version of the new STAN and try it, but im totally unbudgeted now.

I wonder most of all, why did the past STAN version stopped working?

Thanks in advance for your comments

Hi Gabriela,

due to a time restriction in the license of the calcualtion modul Kelly 2011 (used in STAN 2.5.1072) it is not work anymore in 2014.

If you download the new version of STAN (2.5.1103) from www.stan2web.net you will find a new calculation algorithm called IAL-IMPL2013 (it is the old Kelly algorithm with a new license model).

Unfortunately we were only allowed to incorporate a limited version of IAL-IMPL2013 in the new STAN version (total sum of variables + constraints <= 2000). If your model exceeds this limit and you still want to use this algorithm I am afraid you will have to buy a license from IAL (links on our website).

Sorry for not having better news,
Oliver