As I read this thread I am taking a break from writing a 
program that could properly be called a stochastic 
simulation. 

The program simulates a simple small refinery that is
running on an oil rig in the North Sea. Fluids flow from
two fields called Kyle and Banff into this refinery. Each 
fluid is made up (roughly) of three constituents, oil, gas
and water. There are these refining components called
separators that separate the fluids into their constituents,
i.e into oil, gas and water. Once initial separation is done 
the oil from both Kyle and Banff are comingled for further
processing. Same story for the gas and water. But for 
accounting purposes one needs to keep track of how
much oil came from Kyle and how much from Banff. There
is enough instrumentation on board to keep track. 

Similar story for the oil and gas. The gas is however more
complex since part of it is used to fuel machinery, part is
flared off, part goes into a gas transmission line and part
of the gas is re-injected under pressure into wells on the other
end of the field in order to force out more raw fluids. 

The gas is also analyzed into 14 components by gas 
chromatographs. 

Mass, temperature and volume measurements for oil, gas
and water are made at various points through out this 
system so we a can account for where everything goes,
i.e. mass conservation for each constituent and for the
total flows. 

We have built an engineering model in software of the 
various processes involved. Now we driving our model 
with random inputs (normally distribute flows with mean
and standard deviations, a Monte Carlo simulation) in order 
to see what sort of statistics come out the other end of
the beast. Essentially, we have a "what if" and design tool 
for proving up and testing against actual measured results . 

At the end of the day the simulator is just one part of
the actual instrumentation and auditing mechanisms 
used by the owners and various government agencies 
to determine how much oil and gas is coming out of the
hole. Around $200,000/day flows from these particular
fields so it is worth doing the measurements with some 
care. It also costs a lot to change something so it is 
worth a lot of calculation to decide on what to do. 

It is a fun project. Now I find myself hoping the price
of oil goes up, chuckle. Funny how that works. 

HTH,

BobLQ

BTW, this is a consulting job with an oil field services company
that is headquartered in Norway. I do all of the work remotely
via the Internet. One nice perq though they paid most of my 
expenses for one trip to Norway although nothing about the 
work really required that I go there. 

On Friday 30 September 2005 09:37 pm, [EMAIL PROTECTED] wrote:
> I believe 'stochastic' means 'random'.  Many processes you might
> want to mimic on a PC are so complex that,
> practically speaking, there isn't
> an equation to explain it.
> For example, the stock market and the weather.
>
> To simulate these, we simplify our program by assuming they behave
> somewhat randomly or stochastically.....so your physics equations
> will involve random number generators.
>
> Hope that helps.
>
> Chris
>
> On Sat, Oct 01, 2005 at 01:32:12AM +0000, Todd Walton wrote:
> > On 9/30/05, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
> > > stochastic simulation software for the United States Navy.
> >
> > Really...
> >
> > Can you tell me, Chris Seberino, what stochastic simulation software
> > is?  I've only heard the term used in connection with machine
> > learning.  M-W.com, which is woefully inadequate in this context, I
> > know, says that "stochastic" means "random".  But the OED and Google's
> > define: say that it means random, but with some sort of big picture
> > predictability.  Can you elucidate?
> >
> > -todd


-- 
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to