New Standby Generator 1

[candu6]

We are planning to replace one of our standby generators (4.5MW, diesel engine driven) and a consultant has proposed that we replace the single generator with two smaller manchines (2.5MW) which could be supplied as two portable units; then we can easily send them back to a manufacturers service center for maintenance. The single 4.5MW standby generator supplies a distribution bus with a variety of loads. Therefore, two smaller machines would have to be paralleled and share the load.
I'm not crazy about this idea because I don't believe two machines can be as reliable as one and I think the configuration and controls will be overly complicated. Does anyone have experience with a parallel standby diesel generator design?

 

[davidbeach]

Paralleling is not a big deal. With one machine, if the machine is down you have no output; with two the loss of one would still allow 1/2 the load to be carried. Better yet would be three 2.5MW machines so that the loss of a single machine does not cause the loss of any load.

 

[byrdj]

I love it, two out of three, the nuclear way (Have a Star)

 

[davidbeach]

The critical data processing facility way of doing things is load that can be handled by two gens so they have four. When the utility cuts out all four are started. If all four start, after about 5 minutes one is shut down leaving three running to supply a load that two could handle. If one shuts down the fourth unit is restarted. If the load becomes small enough that one generator could handle the load, the third unit is shut down leaving two to carry the load, but they never run in a condition where the loss of a generator would cause load to be dropped.

 

[FreddyNurk]

Davidbeach is spot on. Indeed, in remote Australia, this is exactly what is done with prime power units so that there is always a redundant set running. Granted, the requirements may change a little with the standby application rather than prime power, but its certainly possible.

http://www.daws.com.au/projects/BremerBay.html

shows an example of running up to 4 units in parallel.

I also believe that there are enough options in terms of control systems that allow for multiple sets to operate in parallel without any problems. Woodward is the system I'm most familiar with, but I believe almost any manufacturer would allow this capability.

Note that I have no professional affiliation with any of the above mentioned companies.

 

[itsmoked]

David's idea has many other good points. You have to look at the running life picture too. If you are talking remote-ship-back all your expenses are nasty if you just have two. Think of needing one worked on. You have a 'panic hurry up' situation that will extract larger sums out of the operation to get things back up after the inevitable problem.

I would go for two units then have a third delivered in a year. Cycle thru them all on a regular basis. At Major time you can calmly schedule a ship away rebuild with no down time. If one breaks you can calmly schedule a ship away repair if required. If more power is desired in the operation you can increase one generator in the count. All the parts stay the same so spares are minimal.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[waross]

You say one of your standby sets. Does this set already parallel with other standby sets or does it supply a dedicated load? Do you already have a back-up set for this one?
I like the idea of N+1, that is 3 x 2.5 MW sets. The reason is dependability, not repairability.
These sets typically run 15,000 hours between top end overhauls and 30,000 hours between complete top and bottom overhauls. With the newer onboard computer generated maintainance schedules the time may be extended for a lightly loaded set. It is a lot easier to transport workers than it is to transport megawatt gensets. It is quite possible that the majority of the hours on the set will be scheduled excercise.
In ten or fifteen or twenty years you may be replacing the set because of age and never will have done a 15,000 hour overhaul.
There are some good reasons to use multiple smaller sets rather than one large one, but the ability to drag the set to the service station for an oil change is not one of them.
respectfully

 

[rcw retired EE]

Candu6- A design review of possible savings in the standard CANDU reactor design two years ago determined that several 2.5 MW units were more cost effective and as reliable as one 4.5 MW unit.

The study included a lot of subjective analysis, especially about the cost of nuclear qualified systems. Some larger designs were already qualified but the smaller 2.5 MW units were not. Cost of qualifying testing wiped out any capital cost savings.

If this is for a non-nuclear system, I believe the multiple units will provide the same reliability at lower installed cost. It may be worth investigating.

 

[AusLee]

Both configurations, 2x2.5 and 3x2.5 will work, but i do not recommend it unless your installation is prepared for a loss-of-synchro problem.

This way you must divide your installation in two, which is good if you can do it.

I propose 1x4.5 MVA + 2x 2.25 MVA if you want to have an installation that withstand any kind of problem:

If you lose the big one, the other two can enter in synchro. Probability of loss of sunchro when a generator is broken is << 1.

One of the smaller two can take the load when it is 80% of 2.25 MVA, therefore your running cost will be reduced.

 

[alehman]

AusLee - can you elaborate on the loss-of-synchro concern you mention? Why do you feel it is more of a risk with 2x2.5 and 3x2.5?

rcwilson - I'm interested in the Candu reliability study. Do you know if it has been published?

My experience has been that one larger unit is generally more reliable and less complex than multiple parallel units.

 

[rcw retired EE]

alehman- The report analyzing Candu Reactor standby power system options was not published. It was an internal study done by an engineering firm for AEC. The idea was to look "outside the box" for capital cost and construction time savings in a standard design. 2.5 MW packaged units dispersed around the facility near the critical load centers looked like a good option to a couple of 4.5 MW units. The modern packaged units are almost "plug and play", arriving on site with the fuel tank, controls, switchgear, and weather enclosure all on pre-assembled skid. Total installed cost appeared to be less.

 

[AusLee]

well, like you said, i too feel that the building should have one "principal" back-up generator, so this eliminates all the n x 2.5 MVA possibilities.

Regarding the loss of synchro, i have seen a few examples:
1. the fuel pump in one gen may be responding awkwardly, so it cannot keep same frequency no matter what the excitation, so the faulting generator disconnects automatically from the bus and you have an alarm.

2. the control may diverge, one generator increases kVAR and one absrobs more kVAR until the threshold is reached by one generator which automatically disconnects.

etc.

 

[Hoxton]

Exactly, AusLee, so what if this happens to your only 4.5MVA set? As others have said, if you have a critical load, like a banking centre, multiple sets are an answer. Really, standby sets are an insurance and you have to think in a logical process. 2.5MVA is about the largest genset you can fit in a container, any bigger and you are into site construction (buildings etc.)

So going for 2.5MVA sets, rather than a 4.5MVA does have logic to it.

All applications are different, however.

 

[alehman]

Thanks rcwilson and Auslee. There is a lot to say for staying under the 2.5MW range. Some of my clients aren't willing to pay the higher cost per kW for larger sizes. Other advantages are standardized systems and components and more readily available service and parts. It sounded like Auslee had some particular technical concern with paralleling smaller systems. Mechanical malfunction and VAR control can be problems at any size it seems to me.

My real question is how do we asses the relative reliability of a single generator vs. say a system with 3 parallel units, one being redundant?

It is easy mathematically to compare the relative probability of a load-impacting failure for N systems with R redundant units, if we know mean time between failure of a unit and the mean time to repair. Some reliability data for packaged gensets is available (IEEE 399, etc). But that doesn't account for the controls.

For me the harder part seems to be determining the probability of a control or switchgear failure. I have experiences with failures in two major manufacturer's computerized paralleling systems (in one case, dead bus with all gensets still running). I've not seen any reliability data on such systems.

 

[AusLee]

I have no doubt that currently, the reliability of the control is absolutely zero.

Taking the example of a manufacturer that uses a Microsoft Windows Compact Edition to run a SCADA application, the reliability is that of two items in series: WinCE + SCADA.

Whereas WinCE is harder to crash than the rest of the MS Windows operating system, it is not secure against tampering and/or malformed call procedures. For tampering, the OS might be functional, but an application developed under another environment might overwrite some critical files when installed. For malformed call procedures, it may simply be because the application has written data over the program area due to the fact that the memory is too small on the device.

The worst part is about the SCADA. The tabbed controls, buttons, text boxes, etc. are drag and drop items from an Indegrated Design/Development Environment, and as soon as the program has made the assembly, he has an application. I know a lot of tests performed on financial applications developed for banks; in contrast, i have seen SCADA programmers being called on site to investigate a malfunction, resulting in a mere "bummer" and then repairing the flaw they have identified.

Therefore, until a well established standard for testing control software is available and enforced, the reliability of the control system will continue to be zero = a best effort approach.

 

[waross]

I understand that PLCs and controllers listed by Factories Mutual must meet a high level of reliability and are often a prerequisite to insurance coverage.
Any-one with more knowledge on Factories Mutual, please comment. Do I even have the name right?
respectfully

 

[alehman]

My experience has been primarily from a commissioning viewpoint. We develop functional testing procedures to verify various operating scenarios, relaying, failure modes, etc. after electrical acceptance testing is completed.

A recent horror story involved no fewer than 8 trips by my firm, with the vendor's software guys on site almost continuously for many weeks debugging and uploading new versions of this and that software (including MS Windows). We saw a number of "blue screen of death" incidents during the process. It had a windows front-end with a PLC controller. Our client ran out of money to pay us before we or they were comfortable with the system. Every new software update requires that the testing start over from the beginning. Admittedly, it was a customized version of the vendor's product and was delivered on very short schedule. I felt bad because of the wear on the system as well. Some of the medium-voltage breakers had several hundred operations in the process.

This type of experience is not limited to one system or vendor however.

 

[waross]

Look at this site for reliability testing and forcasting.

http://www.fmglobal.com/about/research/research_reliability.asp

respectfully

 

[advidana]

If it is a standy generator why do you need two of them, twice problems. I know standby generators are needed but a lot of times they don't alway work when you need them. I would stick to one genrator if I was you. Buy what you need. You can alway rent a genrator if yours is in repair.

 

[alehman]

Thanks waross. Several of my clients are use Factory Mutual. I will inquire if they have any data.