Sie sind auf Seite 1von 5

Ruminations on modelling

CF Forster, "Advances in wastewater treatment and pollution control", pp. 204 - 213 in THY Tebbut
(ed.), 1985 "Advances in Water Engineering", Elsevier, ISBN 0 85334 374 8
"Modelling can [...] be used to describe and characterise the various types of reactor used to treat
waste waters. Many of these models have their basis in a combination of the Monod equation, the
microbial growth rate and a mass-balance across the reactor. However, despite the regular, almost
repetitive, appearance of this type of model in the literature, one must question whether they have
any real value in the day-to-day operation of a reactor, in the optimisation of reactor performance or
at the design stage. [...] it is difficult to justify the use of models as an alternative to the proven and
accepted techniques based on the use of loading rates and performance relationships."

How time flies


In 1985 WRc was successfully using the BOD-based ASModel, based around the theories of Gordon
Jones, to optimise aeration systems.
In 1987 the IAWPRC published ASM1, kick-starting the wastewater modelling world. GPS-X 2.0 was
launched in the early 1990s, having dropped the collection of BOD- and COD-based models that
were featured in GPS-X 1.0. Under Imre Takacs GPS-X was turned into the first successful widespread
wastewater simulation program. But expensive! The price in the UK was something like 37,000 in
1993, running under UNIX, with an annual licence fee around 6,000. No licence fee, no GPS-X -
unless you paid for the 'perpetual licence' at around 67,000. WRc produced STOAT in 1994, based
on BOD models, adding COD-based models in Version 2, released in 1995. The list price was 18,000
for a perpetual licence, and 12,000 for UK wastewater treatment companies (Thames Water,
Severn Trent Water, etc.) There was a steep discount rate, taken over from that offered by DHI as
WRc was a DHI partner at the time. Wessex Water purchased GPS-X for an undisclosed price, but
indicated that they got a large discount in return for being the first UK customer. Once there were
two software products competing for sales - STOAT and GPS-X - the list price became a starting point
for negotiation, as customers played off quotes from the two to get a lower purchase price.
Biowin went from a DOS program to a Windows version, and WEST appeared on the scene around
1995 to 1997. Simba briefly flared, becoming possibly the most popular program with wide-spread
uptake in Germany and the Netherlands, before its expansion stalled. Efor, from Denmark, made a
valiant stab, but never broke out of its Kruger client base - eventually it was sold to DHI, who waited
to see if there was a market before developing it - and, since they then switched to WEST,
presumably found that the market was not convincing. Now that MostforWater, the company
behind WEST, has folded, DHI have taken over WEST - but this time they appear to be ensuring that
it will be actively developed. There have been annual updates, gradually extending the features,
coming closer to GPS-X in breadth of processes, Biowin in support of a unified aerobic/anaerobic
model, and providing support for catchments and rivers.
But the market is small. In 2001 all the modelling programs combined reported a total user base of
under 1,000. That contrasts with (admittedly in 2011) a single hydraulics-side company, Bentley,
claiming 9,000 users. If you believe the marketing then DHI is the largest provider of hydraulics-side
software in the water industry, the same claim made by MWH Soft (now Innovyse). I have seen no
figures, only spoken to people at trade shows. I have a soft spot for both DHI and MWH Soft
(Innovyse), having collaborated with both in the past.
Helene Hauduc in her 2011 PhD thesis estimates that the world-wide uptake of wastewater
modellers is 3,000 - 5,000, based on (undisclosed) sales estimates from various software companies -
at a guess, Biowin and WEST. I assume that the lower bound is based on sales, and the upper bound
based on an estimate from her other work of the ratio between users to purchased licences. That is
a large expansion in the number of users - from c. 1,000 after a decade of selling, to c. 3,000 after
two decades. And it probably (Helene does not discuss how she built up her estimates) ignores sales
of SIMBA and (possibly) GPS-X (assuming that it is was Biowin as well as WEST that provided details).
At the time STOAT had not been released as freeware, and its previous base (around 200 or so users)
had shrunk to maybe ten or so active users. (Of course, we don't know how many of Helene's users
are active, rather than historical.) But now STOAT is around 8,000 (May 2017) downloads, and a fair
number seem to be actively using it.

Speed
DHI announces the launch of WEST 2011. One of the claims is that it has a new, faster,
computational core. And now ... WEST 2014. I downloaded, installed, and found that it would not
run in demo mode. I didn't have enough interest to work out why; just that demo mode, with no
licence file, is shut down by the licence manager. The big thing for WEST 2014 appears to be the
release of their version of a supermodel. This should slow WEST down, and it wasn't a leader on
speed. But it does mean that I need to look at doing something similar. Just heard from one
developer for the WEST platform that they will have to update their code, as WEST 2014 is not
backwards compatible.
Update: WEST is now at 2017. I still have not managed to get it to run in the demo mode.
WEST 2009 also claimed to have a new, faster, computational core. There had been many
publications on the techniques used to accelerate the calculations. I had the chance to run some
speed comparisons between this and STOAT at the end of 2010. The results were fairly convincing.
STOAT ran at about twice the speed on a few test cases, all based around a simple activated sludge
unit with a clarifier. I could not work out how to disable the graphing of effluent quality in WEST, so
displayed a comparable graph in STOAT. STOAT with no graphics runs faster still.
I doubt if I will get a chance to run WEST 2011, but I do hope that it will have the speed up - it needs
around 100% to be equivalent to STOAT. I had done timings years back against GPS-X, where the
conclusion was that they were about the same speed. But STOAT and GPS-X have Fortran as the
underlying computational engine (GPS-X uses ACSL as an intermediate language, but this compiles to
Fortran). WEST uses C++. The speed difference can really be regarded as that a good optimising
Fortran compiler outperforms an old (but free) C++ compiler. WEST appears to have used the old,
free, Borland C++, and I think then switched over to OpenWatcom C++. I note that GPS-X appears to
be using gFortran, which has a respectable performance and has the advantage of also being free.
STOAT currently uses Intel Fortran.
And now here we have a comparison of the speed of WEST, GPS-X, Biowin and STOAT, done in
2010, published at WEFTEC by Black & Veatch.

It shows STOAT to be quite fast at calculating steady-state results - interesting, as there is no steady-
state solver in STOAT, so I assume that STOAT was run 'long enough' for results to settle down. It
also shows that STOAT is fast at dynamics. The STOAT model did not include an aerobic digester,
dewatering or sludge liquor returns, so it not quite comparable on time with the other programs.
Dewatering is a fast process in STOAT; liquor returns may or may not slow down STOAT depending
upon the options chosen for handling recycle loops. You have the choice to iterate loops until
convergence, or to carry out no iterations. Iterate until convergence is to be preferred. It used to be
the STOAT default, but since GPS-X chose the alternate approach I changed that in STOAT so that we
were not penalised in naive speed comparisons. The aerobic reactor should have been modelled as a
simple aeration basin and would have provided a small slow-down. But the speed differences are
large enough to suggest that STOAT is fast
However, my own tests of STOAT & GPS-X were that the two of comparable speed - but this is
before GPS-X switched to using Java for the user interface. And, as stated above, that STOAT as
about twice as fast as WEST - which is probably reinforced by the result above. Biowin seems
amazingly slow - which may explain the comments about Dynamita (see below) that their new
program SUMO, is fast. Yes it is, but compared, it would seem, to Biowin. I looked at the Sumo
timings and found them similar to slower than an equivalent in STOAT - but without knowing the
hardware, detailed operational settings, etc., no true comparison can be made. And STOAT was
using ASM1, rather than a supermodel.
The slow Biowin times appear to be a consequence of the use of a supermodel - many state
variables, even when they are not active (such as anaerobic biomass when modelling activated
sludge) and, worse, when the variables take small values it can increase the stiffness of the system.
Plus the Biowin approach solves an ionic equilibrium system, involving an iterative set of equations -
but should usually need only two iterations per time step.
There is a suggestion that the timing study mentioned above should be updated. It may possibly be
modified into a series of tasks, to see how people using the different software packages can produce
designs against a hypothetical case, and compare sizings, performance, outline costings and so on.

Super models
There are now two super models in the open literature. The UCT (University of Cape Town) ASM2 + a
simplified anaerobic digestion model has had the PhD thesis published, while the University of
Arizona has also published a PhD thesis that contains another supermodel. Arizona's is a simpler
version - no biological P removal, no ionic equilibrium, and no messy stoichiometry equations. I hope
to use Arizona's to cut my teeth in this area.

More on WEST
Interesting content from a Kruger employee - that he used to use West, but found the interface
difficult and the modelling approach too academic, so has switched over to Biowin. And, he then
went on to say, Veolia had the same issues and have also dropped West for Biowin. Kruger is owned
by Veolia.
The Kruger man thought that DHI would drop West, but in talking with him it was clear that he didn't
realise the West development team was based in Belgium, not Denmark. Looking at West I don't
think DHI are ready to drop it - I would join with a friend and say that they are developing it towards
the role of high-speed integrated catchment model, with the MIKE family as the slower, higher-
fidelity, tool.
Meanwhile, DHI seem to be pushing WEST more towards a high-speed catchment simulator, with
the MIKE family as the high-fidelity version but much longer runtimes. A US friend sees this as being
the future. Our own simplified catchment model still finds that people are not interested in truly
attempting to unify sewers, rivers and sewage works.

Dynamita
Imre Takacs, the leading light in GPS-X from 2.0 to 5.0, and then with Biowin at Version 3, has
formed a new modelling company, Dynamita.com, and launched a new modelling product, Sumo
(from SUper MOdel, i.e. the single integrated view of the whole biological process, rather than
separate models with conversion processes between aerobic and anaerobic treatment). SUMO has
now been released, and an evaluation/demo copy is available as of May 2017.
The early screenshots show the influence of Biowin on the design. Not much else to report at this
stage. Imre did say that the new simulator would be released for wastewater, but would not be
coupled to wastewater. This sounds a bit like the claims for GPS-X - 'General Purpose Simulator for
X-Windows', which was initially intended to provide a flowsheeting package that could be
customised to any domain. That was in 1992. Twenty years later it still stays firmly wedded to
wastewater treatment. It will be interesting to see if Sumo eventually becomes a generic dynamic
flowsheeting platform. First, it will need to generate significant sales, which I think will mean
persuading existing GPS-X, Biowin and WEST users to switch.
I have no idea what is Sumo's sales pitch to persuade people to make such a move. I don't see the
Hydromantis or WEST users switching just because of Imre's involvement, which leaves Biowin users
as the most likely candidates. I believe that Biowin users make up the largest numbers of the
commercial user packages, so their switch would be interesting. As we get pricing and platform
details from Imre the case for a change may be more convincing, but at present I think it may be a
little too late in a stable market. That said, Biowin overtook GPS-X; WEST crashed, and had to be
rescued. From the news reports on Dynamita's web pages it does look as if people will make the
switch, because of Imre's reputation. The one user conversation reported is a Biowin (and GPS-X)
user, so it is still early to tell if the main hit will be Biowin or a wider market. Going on a few items on
various Linked-In forums the interest is there. A Biowin user I spoke to was interested in details of
SUMO when it was released, to see if it was worth asking if a switch could be made. I don't think she
is dissatisfied with Biowin, but would like to see if a more modern offering would allow her section
to do their work more efficiently, or look at things that they currently cannot. Imre suggested a price
around 2,000, which would make it an attractive alternative.
SUMO has been released. One friend has been given a copy, possibly an evaluation copy. He thinks it
is great, as it answers one of his gripes with current simulators - the blank flowsheet. STOAT and
GPS-X, and I think WEST, come with some customisable built models - but they are not displayed as
the start-up default. SUMO does, and he thinks this makes for improved usability, as engineers want
to get started on solving a problem, not working with the abstracts of building a computer model.
Having been sent a demo copy of SUMO, I can report that you are still faced with a blank screen.
Using a built flowsheet is made readily accessible. The user interface is described as task-oriented, so
that you move from tab to tab and with each tab (choose processes/choose models/configure
data and so on) the menu items change to reflect what is needed for that task. Personally I found
that confusing, back to the DOS days, but I can see that many will like it, and I have used the same
approach for a dedicate GAC simulator - but that designed around a few simple phases. SUMO does
look as if it will be a viable competitor across the range, whether you are looking at Biowin, GPS-X or
WEST.
SIMBA seems to be growing again, with a rewrite to lose the need for Matlab and support from a
Canadian company helping to break out from the German niche. I suspect that at least one of the
two currently dominant companies is suffering, based on some announcements made at WEFTEC -
but this comes from 'reading between the lines', a well-known way of getting things totally wrong.
All of which suggests that SUMO may be coming at the right time.
UPDATE: Three years on, and no one seems to be suffering. But I still think the release of SUMO is
going to affect some of the existing suppliers.
We are not in a zero-sum world - STOAT has, I think, demonstrated that there is still room for
expansion; but I also feel that STOAT has shown that the big market is for something less than
Biowin was charging; what that it, I don't know - all STOAT has shown is that when available for free
many people who ignored wastewater modelling have dabbled. I know of a few who like STOAT
because they can, in principle, give it away free to their own customers - I don't know if they have
done that, though. When we sold STOAT there was always interest in being able to pass STOAT as a
free viewer - which was possible, and, as far as I was concerned, acceptable within our licence. But
none of the people who asked seemed to then go on to deliver STOAT as a viewing platform.
With the exception of STOAT, everyone seems to have a 'model builder' to accept an Excel
spreadsheet that defines the Petersen matrix and spits out the code you need for the new model.
SUMO appears, from comments on the web, to have the most flexible version of a model builder. I'll
have to add such a thing to the STOAT wish-list! WEST and GPS-X both appear to use ASM2d plus the
UCT ADM1 model; SUMO is using something that Imre is developing based on his Biowin experience.
When he and I talk he discusses continuing to tweak the model to better match available data.

Das könnte Ihnen auch gefallen