Today is...
Sunday, May 20, 2018
Welcome to, the global online
community of automation professionals.
Featured Video...
Featured Video
EtherCAT with CTC’s master lets your multivendor network play well together...
Our Advertisers
Help keep our servers running...
Patronize our advertisers!
Visit our Post Archive
3 Element control for boiler drum
What is 3 element control of boiler drum in HRSG?
By Tejas bhatt on 31 July, 2006 - 8:37 am

What is 3 element control of boiler drum in HRSG? What is the criteria for using single, two and 3 element boiler drum control? How does 3 element help in controlling the shrinking and sunking effects? What are the calculations?

Try this link for basic information (it was obtained by Googling "3 element boiler drum level control"):


By Ben Janvier on 9 August, 2006 - 12:50 pm

The engineer I worked with told me about his email so I thought of looking at the replies. Yes, I am the Enero guy who is saying that you should get rid of your steam feedforward (at least filter with a 15 sec. filter) and add a shrink-swell compensator. Of course, you also need to tune the level controller using Lambda tuning techniques (IMC) so that you do not have a level that oscillated continuously.

I will not go over my resume but I will rather cover a few aspects of boilers. If you get a chance to talk to your boiler supplier, you might get the weight of the boiler at no load and the weight of the boiler at high load. Curiously, the boiler weights a lot less at high load than at low load. The reason for this is that the boiler is just like a kettle with steam bubbles UNDER the level. Also, the higher the load the greater the vapor phase above the boiler.

To say that boiler drum level control is a mass balance problem is a mistake. Shinskey, Dukelow and many other "big names" have recognized the fact that drum level control is a volume balance problem and not a mass balance problem.

Conclusion: if you tune your level controller using Lambda tuning and filter the steam flow feedforward you will be doing much better. If you add a shrink-swell compensator you will do even better and will be capable to increase the firing rate of your boiler.

Just a little story. Most of my contracts with drum level controllers are performance based (if the drum level is not 30% more stable the client does not pay) so one day I went to this paper mill and was convinced I was going to be able to improve the operation of the boiler... I was wrong: When the firing rate was increased, the level would swell even with the feedwater flow valve at ZERO; when the firing rate was decreased the level would shrink and the valve at 100% would not prevent the boiler from tripping on a low level.

Note. I have read on one of the replies that drum level is not important. All of the boilers I have seen have a boiler drum level high-low switch in the BMS and the BMS will trip the boiler with the level reaches these limits.

If you have other questions or comments I will be reading the replies.

Take care,

Ben Janvier, President, Senior Control Consultant
Enero Solutions Inc.

Hi Ben,
Yes, the boiler will be lighter at high firing than at low fire.
1. The density of water at say 260 deg C is about 800 kg/m^3 compared with
1000 kg/m^3 at room temperature.

2. As you say, the bubbles throughout the water column will further reduce
the density.

The only way the level of the boiler drum will remain constant over time is
if the total flow of water in to the boiler matches the total flow of water
(steam + blowdown) out - that's why I say it is a flow matching exercise.
Level of water in the drum is not a good measure of the balance because of
the effects of shrink/swell.

It sounds to me as if your "little story" involves a boiler with a grossly undersized drum ...

By Ben Janvier on 18 August, 2006 - 1:56 pm


I have tuned boiler drums ranging from 70 kpph to 1,500 kpph. All of these boilers suffer from the shrink-swell effect but some are worse than the others. The shrink-swell effect is due to the boiler internal circulation in the downcomers. The higher the circulation, the less the amount of bubbles under the boiler. Hence, poor boiler circulation usually translates in long boiler drum level deadtime and high shrink-swell effect.
If you take any boiler (say 70,000 kpph +) and increase the firing rate by 20%, the boiler drum level will go up IF YOU MATCH THE FEEDWATER FLOW WITH THE STEAM FLOW. If you increase the firing by a larger amount, you might trip. My point was that if you want a very responsive boiler drum level, steam feedforward does not do it and you need to take into account the rate of change of the firing rate (on an increase in firing rate, a certain water volume is forced not to enter the boiler through FF logic.)

Ben Janvier, Enero Solutions,

Hi Ben,
All boilers will demonstrate shrink or swell to some extent. The question is whether or not one needs to go away from the "industry standard" 3-term controller in every boiler. There may be some in which mechanical design or exceptional load change requirements or both mean that the response of this system is inadequate, and in these I agree that some further complexity is

Unfortunately, there is a tendency in the industry to use the big guns to shoot everything. We need to remember the KISS principal and not make things are more complicated than they need to be.

My comment was that the water flow and steam flow must match OVER TIME. What this time is will depend on the drum size - but I would expect that it will be less than the hold-up time of the drum - in other words, between 1
and 3 minutes in the typical case. Sure, in the short term, there will be momentary errors in the flow rate, but they must average out to be equal.

Adding feedwater to a drum will usually act to reduce level as the cold water coming in will tend to cool the drum and collapse some of the steam bubbles.

An increase in firing rate which is not accompanied by a load change will increase the boiler pressure. There will be a temporary increase in level due to increased steam formation in the lower tubes, and also a long-term increase because of the effect of increasing saturation temperature on the liquid phase density. If the load change comes first, there will be an increase in steam flow and then a drop in pressure, which will be picked up by the combustion control scheme and used to increased firing to meet the increased load. This is the situation where steam flow feed-forward is of use.

A comment re the "importance" of drum level in one of the earlier posts - I was once involved with a boiler where the client insisted on using 3 level measuring systems for the shut-down - one at each end, and one at the centre. On start-up, we had to deal with the problem that the system at the far end of the drum (away from the burner) was indicating a high level while
that at the near end was showing low. We asked the boiler manufacturer about this and were advised, after he'd done a bit of research, that the circulation of the boiler was responsible - water was rising up at the end away from the burner which saw the full effect of the flame, and was actually flowing down the furnace wall at the burner end.

Solution was to go into the drum and extend the tappings internally so that both "end" systems effectively saw the water level at the middle of the drum.



Yes, I agree the KISS (Keep It Simple Stupid) should be the basis for process control. As such, since steam feedforward will not help us during fast transcient and will destabilize the level controller I am just saying to go to a two-element level controller (which eliminated the steam feedforward and should make you happy since it meets the KISS theory).

HOWEVER, if you are a process control engineer working in a power plant and need to swing your unit faster in order to sell your MW at a premium; or if you are a control engineer working in a paper mill where the boiler firing rate can change quite a bit; or if you work in a chemical plant with wate fuel boilers, I challenge you to go and see the plant manager and say: "Well sir I chose the KISS theory and decided that it was not worth it to make our boilers swing faster".

Sometimes however, the boiler firing rate is constant and there is no need for feedforward and the KISS theory works like a charm and only good PI tuning is necessary.

If you need a fast boiler forget about steam feedforward and just program a firing rate feedforward.
If your boiler does not need swing than KIS (just one S) and a shrink-swell compensator is not required.

Hope this helps "clarify" my position.

Ben Janvier, Senior Process Control Consultant


Steam feedforward is more direct in the event of a steam flow demand change. I would agree that firing feedforward is possibly better if a load change is initiaited from the firing side. However, in the event of a steam demand increase, the steam flow will increase - this will cause the pressure to commence ramping down - the fall in pressure will initiate an increase in firing demand - and eventually the firing will ramp up after the air has done its thing. It's easier to slow down an excessive rate of demand than to speed things up.

I think we are going to have to agree to differ on this - but I have seen a boiler handle a sudden draw-off of steam into a cold header without losing level, with steam flow feed-forward. The boiler was firing at about 30% into a vent during a shut-down when someone accidentally opened the isolating valve on to the cold header - steam flow went up to 100 % over about 20 seconds.


I appreciate your post and working in the Chemical industry, feedforward control (or single element) is completely insufficient to handle large swings in Fuel Gas specific gravity or steam pressure changes.

We operate two of three B&W boilers (450klb/hr, 600# steam capacity). We supply steam to an Olefins plant and several other derivative units. The olefins plant generates 1500# steam which drives turbines that have 600# extraction. Tail gas from the olefins unit feeds the plant fuel system which means that the boilers can see anything from H2 rich to fuel gas with ethane in it.

We have had numerous problems as you have noted from your paper plant experience with the boilers catching even with three element control before reaching the low level trip point. I'd be interested in talking with you regarding your experience.

By Walt Boyes on 19 August, 2006 - 2:25 pm

There is an excellent article on drum level control in the August issue of _Control_, available at no charge on The article is by David W. Spitzer, and has input from F. Greg Shinskey and other significant authorities on boiler control.

Walt Boyes
Editor in Chief
CONTROL magazine
Putman Media Inc.
555 W. Pierce Rd. Suite 301
Itasca, IL 60143
blog: Sound Off!! at
630-467-1301 x 368

By Mehmet Mercangoz on 17 August, 2006 - 11:44 pm


Could you please explain what you mean by a shrink swell compansator?

Best regards,


By Ben Janvier on 19 August, 2006 - 2:27 pm

A shrink-swell compensator will take the firing rate as an input variable, filters that variable (say 3 seconds filter) and then applies a d/dt block (rate of change or derivative) to it. What you end up with is a firing rate of change calculation.

You multiply the rate of change by a negative gain and this is your shrink-swell compensator.
Firing Rate is constant so rate of change is zero and no change on feedwater flow
Firing Rate goes up and rate of change is positive, multiply by a negative gain forces the feedwater flow to drop to correct for the level swell.
Firing Rate goes down and rate of change is negative, multiply by a negative gain forces the feedwater flow to increase to correct for the level shrink.

This Feedforward scheme has been applied many times and is allows faster boiler response.

Ben Janvier, Senior Process Consultant

By The Boiler Controller on 5 August, 2006 - 10:20 pm

1- element boiler drum level controller:
Level is controlled using as a simple feedback loop.

2- element boiler drum level controller:
Level is controlled using a cascade configuration where the drum level PI controller send s flow setpoint to an inner feedwater flow controller.

3- element boiler drum level controller:
Same thing as two-element but this time the steam flow is used as a feedforward signal. When the steam flow goes up the feedwater flow setpoint is increased automatically.

How does 3 element help in controlling the shrinking and sunking effects?
Re: It does not help. To the contrary it makes it worse. For instance, when the firing rate of the HRSG is increased, the steam flow goes up and the level swells. The problem with steam FF is that the steam flow controller will force the feedwater valve to open while the level controller will be on its knees begging for less water!

If you have a boiler or HRSG that necessitates faster firing rate and you are stuck with the shrink-swell effect, you should eliminate the steam feedforward and add a firing rate feedforward logic. I had some consultants come in my plant and the max firing rate limit was increased by two and the drum level is now more stable.

If you are insterested in boiler controls, you might want to look at this course
(course given by these consultants):

Hope this helps,
The Boiler Controller

I can't let you get away with that! The boiler drum control problem is not a level control problem - the objective is to match the feedwater flow to the load demand on the boiler. The need to hold drum level is a secondary requirement. The shrink/swell effect is a short-term one that will last maybe 20-seconds-1 minute depending on the size of the boiler - after that the effect will have gone away and the expected direct integrating response will appear in the level. The drum level controller should be set up so that acts as a trim controller with comparatively low gain (otherwise the shrink/swell will affect the feedwater demand) and a low integral action - enough to restore the level to its set-point over time but not enough to cause instability. The three-element system effectively simply gives a derivative component since the flow difference is proportional to the rate of change of level. Your proposed firing rate feedforward will not react fast enough to a sudden demand in steam load such as adding an extra load to the system - the drum pressure has to fall to increase the firing rate, then the combustion control system has to respond to this to increase the firing rate, before the feed-forward cuts in.


By Phil Corso, PE on 6 August, 2006 - 7:26 pm
1 out of 1 members thought this post was helpful...

Responding to Bruce Durdle's Aug 6 comment about "recovery" from the
shrink/swell phenomenon being dependent on boiler size.

I believe the phenomena is more likely impacted by steam-drum size, not
boiler size (capacity!)

Phil Corso, PE {Boca Raton, FL, USA}
[] (

You're probably right - but since most drums are sized to give about a 10-minute hold-up time, the two are often directly related. (At one time I worked for a petrochemical firm who had specified an extra-large drum on the waste-heat recovery water system on the reformer to give better protection against boiler trips due to level problems - but we still ended up with a 10-minute hold-up time because the boiler manufacturer simply put the level tapping points closer together.) Actually re-reading what I sent, what I meant was that the swell/shrink effect (less letters than "phenomenon") will last longer in a larger (higher) boiler because it will take longer for the bubbles formed in the water wall tubes to rise to the top and come out of the system - but I may be wrong. Anyone got any hard information on the dynamics of swell or shrink? There is another good word associated with these effects - "Squash" is what happens when the operator panics on seeing a low drum level, puts the feedwater system into manual, and adds a LOT of COLD feedwater in a hurry. Guaranteed to drop the water level below the sight-glass.


Dear Bruce,

Good day to you! From the web I know you are also intersted in the boiler drum water level control. We have develped the most accurate boiler control system in the world. You can browse it from the website:

You can find how we can control the level error under 30mm in supercritic drums with five kinds of water level gauges.

Waiting for your reply and comment!

By K Karthikeyan on 11 September, 2006 - 10:49 pm

Hi Dean Wang,

I noticed some interesting item in your site
The "Internal Electrode Sensor of Water Level of Steam Drum (ZL2004.2.0073417.2)" appears to be mounted internal to the drum.

I have some questions:
1) What is the advantage of mounting electrode internal to drum
2) How the electrode cables are taken out of drum? 3) How to replace a faulty electrode during operation?

Also the "High Precision Sampling Electrode Sensor of Water Level of Steam Drum (ZL 00 2 21067.6)" I have some qustions.

a) what is level difference in drum and sensor vessel
b) why sensor electrode pitching distance is varying instea of constant?
c) what is water quality required to achieve 3 yr electrode life?

i am reachable at


By Allan Evans on 6 August, 2006 - 10:19 am

I am a lead Controls & Instrumentation commissioning engineer as well as the plant start up engineer. I am in my 39th year of commissioning power stations, large industrial boilers including HRSG's. From time to time I read the Q & A on this site and wonder how my peers actually get their boilers to work. The answer to this one is a classic case.

Three element control, tuned correctly, with no fancy add ons works very well. Two element control is actually steam flow trimmed from the drum level.

I wonder just how much silly information is passed on by engineers calling themselves "consultants"

Allan Evans.
New Zealand

By Martin Young on 9 September, 2006 - 12:14 am

I just wonder why so many replies talk about firing rates and BMS when the original question was regarding drum level control of an HRSG.

I would like to know more about 3 element control.If anybody can help me pls mail to me. thanks