Different Size Pumps in Parallel 2

[mls1]

I've recently integrated a new pump at a water treatment plant that is in addition to two existing pumps. All three pumps are VFD driven with speed control from a PID loop in a local PLC controlling discharge pressure at a constant value. The existing pumps ramped together and worked well but the new pump is twice as big so it really can't ramp at the same speed. So we operate in two general modes:

1.--The smaller pumps coming on first and ramp to maintain pressure. If they reach 100% and can not maintain pressure, they are fixed at 100% and the big pump is started with the PID output assigned to it.

2.--The big pump comes on first and ramps. If it gets to 100% then it is fixed at 100% and the small pumps come on and ramp with the PID output.

When switching the PID from small pumps to big and vice versa I manually correct the control variable to smooth out the transition. So far the system is stable and the transitions work well.

Here is the problem, if the plant is in mode 1 and the operators want to switch to mode 2 or vice versa there is a serious pressure transient while the pumps adjust to the new mode. I can see several ways to try to smooth this out but most are pure brute force. Has anyone else had experience with a system like this and what was your solution?

 

[BigInch]

Don't wait to get to 100% speed with one group before coming on with the other.

http://virtualpipeline.spaces.msn.com

 

[mls1]

I thought of a possible solution along those lines that would operate all three pumps at variable speed at the same time. If I use the pump equations and scale the big pumps speed signal so that its shutoff head is about equal to the other two pumps and then let them ramp together, granted they won't have the same characteristic curves, but they should at least share the load such that no pump is shutoff. The small pumps will hit 100% first and then the big pump will pick up the lion's share as it continues to ramp up. Would that be a better solution?

 

[BigInch]

I would put them all on discharge pressure feedback speed control, unless you are starting or stopping any group of pumps. You should then revert to a programmed start/stop sequence override.

Starting up ideally you would have a recirculation line from individual discharge back to individual suction, with any starting pump blocked off from the header (with a discharge check downstream of the recirculation line) until it reached its speed setting corresponding to the current discharge pressure of the group, but it may be too late for recirculation lines.

If you can't block off a starting unit from the header while it is ramping up using a discharge block valve and an open pump recirculation line, the starting pump will see the header pressure as a static head that it must equal before it starts making flow into the header. In effect, from the starting unit's perspective, you have a "virtual" block valve. From the header's perspective, you have a downstream branch into which the header will send all flow possible, until the starting unit finally reaches the header pressure, when it will immediately start dumping its flow (corresponding to speed) into the header. That crash of flows is what (probably) causes the hammering.

Until the ramping unit reaches header pressure, the flow corresponding to the differential head it is generating at its current ramping speed will recirculate internally within the pump casing at low efficiency and generating heat.

To minimize the possibility of hammering when starting (or stopping) you should first try to reduce the speed of the running group of pumps to as low a speed as possible, that group making less flow for a short time, as you ramp up the starting group of pumps. As the starting group reaches the current header discharge pressure, you could then further reduce the speed of the running group (now the stopping group) and increase the speed of the starting group. Once the starting group overtakes the current (average) differential head being delivererd to the header by all pumps, ramp down the stopping group as quickly as possible and continue ramping up the starting group. If you have discharge block valves, the time to open the starting unit's block valves and close the stopping unit's block valves would be whenever the discharge head of all units was equal.

I've got transient analysis capable software that can simulate a controlled starting/stopping process, so let me know if you're interested in optimizing those sequences.

http://virtualpipeline.spaces.msn.com

 

[Valvecrazy]

Pumps in parallel can be tricky to control. As the flow is reduced, whichever pump builds the most head is doing the entire job while the other pump or pumps are being deadheaded. The pressure transients you are seeing are caused by the slow reaction of the drives. Pumping water is like pumping a rubberband. When the demand increases, it will stretch the rubberband because the pumps do not react instantly. As the pumps ramp up to catch up with demand, the stretched rubberband finally snaps back to normal which causes a transient. When the demand is reduced, the pressure builds up and the rubberband is compacted until the pumps finally slow down, causing another transient.

The only way to stop these transients is for the pump controls to instantly and exactly match the demand. Setting a drive to react this fast causes it to continually hunt instead of steady out.

I have found that a constant pressure pump control valve on each pump can stop the transients. It also keeps any pump from having to buck the pressure of any other pump. Since each pump has it’s own control valve, each pump is only bucking it’s own back pressure, not that of the other pumps. These constant pressure valves (CPV), can never completely close. The seat is designed to allow 5 GPM to pass, even when the valve is in the closed position. This keeps the pump cool without the need for a recirculation line. Since the valve never completely closes, it is also designed to react almost instantaneously. This high speed reaction, would not be possible with fully closing type valves, without causing water hammer or transients.

When maintaining a constant head or pressure by throttling a constant speed pump, the energy used by most pumps is almost identical to the energy used when the pump is slowed with a drive.

There are several ways to set up this type control. My favorite would be to stagger the pressures so that the smallest pump runs first and the largest last. Say for instance you require a minimum of 60 PSI. I would set the largest pump as pump 3 and have it come on at 60 and go off at 65 with a standard pressure switch. Pump 2 would come on 65 and off at 70. While pump #1 would come on at 70 and only go off at 75 if the demand ever got lower than 5 GPM. Each pump would have it’s own CPV, which would be set to hold a downstream pressure the same as that particular pump started.

With across the line starters, when the demand increases, these pumps will instantly come on and instantly supply the correct amount of water needed. This keeps the rubberband from stretching as an increase in demand is instantly met with the correct increase in supply. When the demand decreases, these valves instantly shut down to match the new demand. This keeps the rubberband from compacting which causes pressure spikes. If the system pressure increases by 5 PSI, the lower pressure pump is shut down as it is no longer needed.

All this is done while each pump is guaranteed a minimum of 5 GPM for cooling purposes. I have systems with more than 10 pumps in parallel that are controlled this way and do not produce pressure transients on start up or shut down. Of course this type system only works when pumping fairly cool and fairly clean water only. Pumping any other kind of liquid or even water at high temperatures is not possible with this type control.

 

[BigInch]

valvecrazy,

wouldn't control valves on each pump discharge, when each pump is already fitted with VFD, kinda' an overkill situation?

http://virtualpipeline.spaces.msn.com

 

[mls1]

Thanks, all. You've pretty much confirmed my suspicion that this is not a trivial solution. BigInch is correct that it is too late for recirculation lines and that throttling is undesirable since the plant is very much looking to control energy costs.

Valvecrazy describes my conundrum well. I can make it respond very quickly to reduce the magnitude of the transients but then I risk making the system unstable. But at the same time this system is a city water supply without stored energy backup (i.e. no water tower) so wild fluctuations in pressure cause fire alarms throughout the city, not good.

As it is now, I have it quite stable as long as it is kept in one mode. It's when the operators switch modes that we have trouble. The best short term option may be to just train the operators to take manual control of all pumps, manually perform the switch and then put it back into auto. I'll need to make sure the control variable is at or near the manual level before auto is reestablished but as long as the operators can handle the shift it should work.

 

[Valvecrazy]

I was talking about getting rid of the drives and going to across the line. The only way the drives will react fast enough is if you turn them into soft start only, and give them 2 seconds or less to get to full speed. Then the control valves can react fast enough to keep up with demand, and the transients and low pressure alarms will go away.

When supplying something like a city without elevated storage, it is imperative that a slight decrease in system pressure be instantly met with an increase in flow. It is just as important that a slight increase in system pressure be instantly met with a decrease in flow. Any hesitation in response, which drives are programmed to do, will cause transient pressure waves. I had always thought that soft starts and delayed reaction helped eliminate water hammer. After studying some computer generated graphs of transients, I can easily see now why instant reactions to changes in pressure have helped me eliminate water hammer in many systems. Contrary to what many believe, soft starts and slow reactions actually cause water hammer instead of stopping it.

I can give you a few references of cities that have changed from drives to constant pressure valves, and have almost completely eliminated all line breaks. I have supplied cities of up to 40,000 people with a single 80 gallon bladder tank. As long as the pumps can instantly and exactly match the demand, pressurized or elevated storage is not needed, and the transients will go away.

Anytime you leave a system to manual control, you are just asking for trouble. Even if the people there can do it right, those same people may not be there the next time.

Lastly; the biggest misconception going is that throttling waste energy and drives save energy. If your pump has good brake horse power characteristics, as most do, energy consumption at low flow will be virtually the same no matter if the pump is restricted with a valve, or slowed down with a drive.

Just get out your pump curves and look at the power required when the drive is spinning the pump at the lowest possible speed that will produce the pressure or head required. Then look at the same curve when the pump is at full speed but, restricted to the same flow rate. Even if the difference is more than 5%, when you add back in the parasitic loses of the drive itself, there will be very little if any energy savings from using the drive. If the drive can show any energy savings, you have chosen the wrong pump in most cases.

The fact that you lose head by the square of the speed, keeps you from being able to slow a pump enough to save any energy when constant head is required.

 

[BigInch]

I totally agree with everything you are saying, especially about the false sense of energy savings when using VFDs, except for your second paragraph. Slow acting pumps are not likely to cause transients, although they may amplify any transients that do appear in a system.

I think the transients that you had there were produced by rapidly varying demands downstream and your valves acted fast enough to counter those waves as they reached the pump station. It is possible that valves can react fast enough to fill in the bulk of the troughs and clip the crests of a transient wave, where a slower acting drive response would not have a chance to do so before the transients hit the speed controls and, by the time the pumps reacted, the speed adjustments were in phase with the transient and exacerbated the situation by actually increasing the crests and deepening the troughs of those waves. A transient analysis does little good if it is not modeled with exact control equipment response times and pump moment of inertias. For example, if the program default moment of inertias are used, which are usually way too small and, thus give a pump a false capability to change speeds rapidly and counter fast acting transient waves, whereas in reality they actually act much slower and can amplify the transients if they do anything at all.

http://virtualpipeline.spaces.msn.com

 

[Valvecrazy]

Biginch you are right. The transients are already there. The slow reaction of variable speed drives just amplifies it. Perfect explanation!! Modeling with the defaults makes drives look much better than what happens in the real world.

It is counter intuitive that throttling a pump can make it work easier. It is already in everyone's head that slowing the speed can save energy, and choking back a pump should make it work harder. I am just amazed at the number of educated people who do not understand this is false. There are countless articles in the trade magazines that falsely proclaim the energy savings of variable speed. Certain states require variable speeds to limit energy consumption above a certain horse power. There are even electric companies who give discounts for the installation of a drive. There are very few people who understand it well enough to be willing to stand up and say it is all a hoax.

In reality anytime you vary the speed of a pump you are using more energy per gallon than when running at BEP. The only energy savings of a drive is to eliminate the end rush currents on start up. Not only can a simply soft start, even auto transformer, accomplish this same task but, when varying the flow to match the demand, there should be few starts anyway.

There are lots of other good applications for drives. I don't understand why it is so important for some people to falsely include pumping cool water at constant head in that category.

 

[ccfowler]

mls1,

Stability can be a serious issue with parallel pumps if they do not ALL have sufficiently steep head vs. flow characteristic curves. ALL pumps must have head vs. flow curves that rise continuously to shut-off. If your pumps have somewhat flat curves, the adjustable speed drives may be contributing to instability problems.

BigInch and Valvecrazy are right in recommending that this is probably not a good application for adjustable speed drives. It is quite likely that the energy consumption of your system is worse with the adjustable speed drives than it would be if the motors were simply operating across the line.

It is likely that your most favorable operating arrangement would be to operate with the lowest practical system pressure.

 

[BigInch]

ccfowler,

I don't think I said its not a good application for VFD in this case. Normally I don't like VFDs, but there are applications where they do work and work well. Its just that I can't tell in this situation because I don't know the degree of variation in flowrate, something I consider essential to decide if VFD's work or not, so I don't know.

Valvecrazy,

Yes, I have noticed for a long time that the pump world is full of false claims of energy savings and waterhammer solutions using VFDs. Sometimes I think its my mission in life to lay bare the truth, but there is so much false propaganda out there its a difficult job trying to squash it all. The propaganda never mentions the reduction in head that comes with the reduction in speed, even making recommending a variable speed drive for a submersible deep well pump! #!Arg~&#

http://virtualpipeline.spaces.msn.com

 

[quark]

I think I am missing a point here. the OP says he doesn't have problem with either of the modes and the ramping happens smoothly. But the problem starts when the operator switches the mode. It is not clear whether the mode selection is done while the pumps are in operation or not.

If the mode selection is done, from 1 to 2, when the pumps are running, the small pumps already running at 100% speed should be stopped immediately as they are to be ramped as per control logic of mode2. Same is the case with big pump in case of switching from mode2 to mode1. Can it be that hammering is caused due to this reason?

ccfowler,

Can you please explain how a flat curve pump can be unstable by a VFD control? The pump when left on its own can have varying flowrates in a very narrow band of head depending upon how flat the curve is. But when you use a VFD, the head can vary by any amount and it should follow only the affinity laws (say, approximately) and the pump curve behaviour will be suppressed.

BigInch and ValveCrazy,

I always consider VFDs when the pump has to operate on system curve and no VFDs when the system has to operate on pump curve. One advantage I experienced with variable flow pumping systems that are recirculating in nature (chillers, cooling towers etc) is that VFDs take care of piping design redundancies (higher friction factors, consideration for aging of the pipelines, future requirements and unconventional pipesizes etc). However, I never bothered to fill a tank with any pump.

I see the OPs application here as a suitable one for VFDs not interms of energy savings but to maintain constant header pressure. Nevertheless, I am a novice in transient analysis. I will request your explanation once the OP answers my question.

 

[BigInch]

I think that's why they are getting transients. Seems to be starting the second group while the first is at 100%.

Instability from flat curves causes the speed control to hunt the other pumps. Flat curves intersecting the system curve is the unstable part. Feedback doesn't work, because there is no dH/Q to work with, both H1 and H2 are equal when its a flat curve so dH/dQ = 0

Pumps in parallel MUST always be producing the same discharge pressure into the header (assuming no significant pressure losses between pump discharge connection points into the header), in order for the pump to flow. The pump does follow the affinity laws, but still it cannot produce a net flowrate into a header, if the header has a higher pressure than it can deliver with its current rpm setting. A pump producing 99 psi into a 100 psi header is a sink taking a backflow at a 1 psi differential from the header into that individual pump's discharge piping (unless that pump has a discharge check valve between it and the header, in which case, the check valve would be theoretically completely closed with 100 psi on the header side and 99 psi on the pump side). If the check is not there, there is basically a virtual check valve there as far as flow into the header from the pump is concerned, but that virtual check allows backflow from the header into the pump until the pump discharge reaches header pressure.

Actually one of the best applications of a VFD, when they should be applied at all, is to compensate for flowrates that have changed since the system was designed and modifying the existing equipment and/or configuration or adding a tank, etc. is not practical for one reason or another.

http://virtualpipeline.spaces.msn.com

 

[Valvecrazy]

Biginch, you are right on the money. There is so much false propaganda or hype about the capabilities of variable speed, that almost everyone has already been completely brain washed. I have written countless letters to the editors of magazines and even the Hydraulic Institute about all the false energy savings claims in articles about variable speeds. They refuse to print my letters, saying that I am entitled to my opinion but, they disagree with me. I tell them that it is not just my opinion, it is an easily provable fact to anyone who can read a pump curve. Now I do not understand if these educated persons really do not understand how pumps work, or if there is just so much money spent to perpetuate the lie about variable speeds, that these editors just do not want to say anything against their major advertisers. Either way they usually just stop replying to me when they are unable to prove that variable speeds save energy.

I guess the function of pumps and variable speeds is just so complicated to some people, that even the experts at the Hydraulic Institute do not understand how they work. Either that or the bigger the lie, the easier it is to get people to believe it. I recently talked to an engineer for PG&E who understands that variable speed does not save any energy. Even he thinks that the VFD propaganda is so strong that is almost impossible to get people to see the truth. There are some major companies who should be in big trouble for false advertising. However, most people are so confused about the subject, that they adamantly believe the hype. I think that Biginch is one of only about six people in the world who understand this subject well enough to render a correct verdict. I think that 4 of these 6 people are also just afraid to stand up and say, "the Emperor is not wearing any clothes".

Quark, it may look as though the drives are ramping smoothly and the transient problems only occur when switching modes. However, even when ramping smoothly, by the time a drive has sensed a difference in pressure and changed the speed of the pump, the transients have already been amplified. It is just harder to see when you are in a mode than when changing modes. Also, a drive cannot increase the maximum speed of a motor, or overloading quickly occurs. Therefore, you cannot make up for increases in head (higher friction loses, future requirements, etc.) without increasing the horse power of the pump. As long as the pump was designed large enough to handle the additional head, the pressure setting of a pump control valve can be increased as well, and will do as good of a job as a VFD in varying the flow.

Again, any slight delay in response, which drives are designed and programmed to do, will amplify the transients even when it looks like it is ramping smoothly.

We haven't even talked about all the negative side effects of drives such as harmonics, voltage spikes, stray voltage, EDM currents in ball bearings, resonance frequencies, reflective waves, etc. If you add in the short life expectancy, lack of repair ability, high cost, continued maintenance, and the damage drives do to motors, to the fact that they do not save energy, then drives can actually increase the cost of operations.

 

[EdStainless]

Just a few practical comments. All of my experience is with multi stage centrifugals, usually 50 or more stages, either downhole or as high pressure surface units.

We only used VFD in a few cases.
-When you are forced to work with a wide range or unknown flow (like well testing).
-When you need a specific flow/head that falls between pump sizes (fixed off-speed)
-and most commonly, when you want flow/head that is above the standard rating (120%).

There are good VFDs, and not so good. A poor VFD will eat both power and motors. In some cases an old motor-generator set would be a better option.

The only justification to using a VFD is to keep the pump running near BEP. If the pump would run at acceptable conditions without VFD then don't do it.

We never put multiple pumps in parallel all on VFDs. Usually the larger pump will have a flatter curve, run it at line speed. Then the smaller pump with a steeper curve can be put on VFD to handle control issues.

I have walked away from jobs where they insisted on using three identical pumps and wanted to handle a wide range of flow.

In your case you have pumps rated at 25%-25%-50% right? What is your operating range?
Is it ever below 25% or above 75%?
Maybe you should just have one small pump on VFD and run the others fixed speed. They will be more efficent and your hardware will last longer.

= = = = = = = = = = = = = = = = = = = =
Rust never sleeps
Neither should your protection

http://www.trent-tube.com/contact/Tech_Assist.cfm

 

[BigInch]

In every case I've looked at, it takes quite a bit of work to optimize both a parallel pump configuration and the operation sequences. I wouldn't have the patience to do it without a good transient simulation program capable of accurately modeling the controls.

http://virtualpipeline.spaces.msn.com

 

[Valvecrazy]

I usually get to do my modeling in the field. I get handed a hydraulic nightmare and am told to do the best I can. No matter the size or the kinds of pumps I get handed, I have had excellent luck controlling pumps in parallel by using a constant pressure valve on each pump. As long as the set point of the valves are all lower than the deadhead pressure of the lowest head pump, each pump is completely isolated from the other pumps. The pump and valve with the highest set point gets its full share of the load before any other pump is started. The pump and valve with the lowest set point, only contributes the extra amount needed over what the other pump or pumps is already producing. The non closing feature of these valves keeps any of the pumps from ever being in a deadhead condition.

I have used this type system with submersibles that are in different locations, centrifugals in close proximity, vertical turbines that are close together or far apart, and combinations of all the above. Two to five pumps in parallel are easily done this way, while I have done as many as 12 pumps in parallel with a little extra planning.

As I said earlier, no pump is ever in a deadhead condition, and the fast reaction of the control valves virtually eliminates transient problems. I had pure xxxx trying to run pumps in parallel with VFD. That is what got me started using valves, and now I am crazy about them.

I have replaced hundreds of existing variable speed drives with valves in the last 15 or so years. The people who are responsible for the operation of the water systems are the ones who really like it and can easily see the difference.

 

[mls1]

Thanks everyone for the excellent posts. Before we started this work I requested a meeting to discuss my concerns. As it is, I am just the integrator not the designer so I can only do what is specified. Now that I've integrated it many of the problems I predicted have come to fruition. I'm trying to nip this before the finger pointing starts so I've called for a "let's fix this meeting". Hopefully we'll get the right attention and actually discuss solutions that work. Your comments will be very helpful. I'll follow up with how things go. Thanks!

 

[BigInch]

Valvecrazy, I love it. Now I understand the reason for your alias. Keep up the good work.

mls. I've been wanting to so some fresh parallel pump control simulations. If you want to send me your curves and basic system details, I'll use it as a base model and let you know whatever interesting(?) things come up. Please find my e-mail on my msn space pages, under contacts, if you wouldn't mind me working with your data.

http://virtualpipeline.spaces.msn.com