Improving drum level response in industrial boilers

Noodnik wrote:

I don't really do process control at all, so I can't recommend literature.

But I can ask questions!

Have you analyzed the current system to make sure that it is harmonized?
  I.e. is everything in the system limiting it's response speed about
the same amount?  If there's some salient part of the system -- a
sensor, or a valve, or something -- that just can't keep up, then fixing
that may allow you to tune the whole thing for a higher bandwidth and
therefore go faster.

Ditto if there are nonlinearities or pure delays in the system that are
being handled by de-tuning a linear control rule rather than by
introducing 'smart' nonlinearities into the controller.

Literature about this would (in my curmudgeonly view) fall into two
categories.  The first will be those that address the questions above,
in a manner that is specific to your problem (i.e. discussing specific
actuator and sensor technologies, expected nonlinearities, etc.).  The
second would be documents written by some egghead in a university who
had no clue how the real world was put together.

I'm interested in seeing if there's someone out there who can answer
your _real_ question.

--
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com

I have had some experience at this from a long time ago when I was in
the navy.
My first question is are there bigger variations in demand than when
the boilers were first installed?
The PID control will only respond so quickly.
Are the feed forwards right set right?  At a constant load the PID
should cause the level to reach the set point. If the feed forwards
are setup to match the flow demand the PID should be contributing
little at this point.  You need to test this at two power level and
maintain them to get the feed forwards gains right.

I would look at the webpages here.  3 methods of control are listed
here.
http://www.controlguru.com/wp/p48.html

I was involved in power generation over 30 years ago.  I know of the
problem but are probably better ways to solve it.  Back then we only
had mag-amps.

Peter Nachtwey

On Sat, 01 May 2010 13:34:27 +0800, Noodnik wrote:


Your control scheme does not take "swell" into account, and that needs to
be done for optimal boiler drum level control.  I don't know of any
literature offhand, try searching for "boiler swell compensation" or
similar.

The swell effect is the increase in steam fraction on increasing firing
rate, which causes the water level to rise and produce an error signal to
decrease feedwater flow, partially negating the steam flow signal to
increase feedwater flow and tending to cause overshoot on the
correction.  If you produce your error signal not directly from drum
level but from drum level plus a leaky (decaying with time) swell
predictor, you can eliminate the undershoot in the response.  Of course
decreasing demand has the opposite effect; lower steam fraction causes
the water level to drop while feedwater demand is also dropping.

It is not possible to eliminate boiler swell, it is only possible to
prevent it from winding up your PID error signal and causing overshoot,
which is usually the biggest cause of poor water level control on boiler
load change.

Thanks, and to the other respondents. The main disturbance here is dips in
HP header pressure, causing sudden increases in steam rate for all the
boilers on the header. Swell occurs also in this situation. We'll try your
suggestion of a leaky predictor (haven't come across that terminology
before, nice) on a simulator and if it looks OK give it a try at the sharp
end.

You mentioned a 3-element boiler steam-drum level control with feed-forward from
steam
rate.  This is not the way that such a control is supposed to work (i.e.
feed-forward); instead the measurement of mass-flow of steam is meant to set
directly
the set-point for the mass flow of make-up water, the PID control on level is
then
supposed to make a slow additional correction for errors in the mass-balance
measurements. Provided the steam:water mass-balance is maintained then there can
be no
better control, and precise control of level is not necessary or practical. The
swell-capacity of the drum must be capable of handling the swell due to
transient
pressure changes, and that's a drum design issue which, if wrong, cannot be
improved
by fancy control.

I have seen many properly-designed traditional 3-element feedwater control
systems
replaced by systems designed by process-control engineers without specific
boiler-control know-how, in which the steam:water mass-balance performance has
been
destroyed by downgrading the steam measurement to a mere feed-forward onto the
make-up
flow, with consequent frequent trips.  It may be worth checking this hasn't
happened
to you.