99.999% Reliability Calculations

[sberger3]

Long time reader, first time to post. I could not find any posts for this subject, so here goes.

We are attempting to design an electrical system that will achieve 99.999% reliability at the load. The only way I was able to achieve that much reliability through SKM software, was modeling a dual fed panel feeding a dual feed VFD. I am not aware of any dual fed VFDs or dual feed panels on the market. Does anyone have any suggestions?(transfer switches do not seem to help: transfer switches supply an alternate feed. The model is requiring two simultaneous feeds to the load, just like data centers have in their dual feed servers.)
thank you in advance,
Sheldon

 

[rbalex]

Long time reader, first time to post. I could not find any posts for this subject, so here goes.

We are attempting to design an electrical system that will achieve 99.999% reliability at the load. The only way I was able to achieve that much reliability through SKM software, was modeling a dual fed panel feeding a dual feed VFD. I am not aware of any dual fed VFDs or dual feed panels on the market. Does anyone have any suggestions?(transfer switches do not seem to help: transfer switches supply an alternate feed. The model is requiring two simultaneous feeds to the load, just like data centers have in their dual feed servers.)
thank you in advance,
Sheldon

What is the nature of the load being served?

 

[sberger3]

Battery charger for controls of a generator tie gear.
Thanks for your reply.
sheldon

 

[rbalex]

I'm confused; what is the purpose of the VFDs?

 

[petersonra]

I'm confused; what is the purpose of the VFDs?

putting VFDs into the system is not a good way to increase reliability.

99.999% up time is only 5 minutes per year of down time.

 

[sberger3]

I gave you the easiest model with the generator tie gear.The charger is for the batteries for the generator tie gear switching.

The second model is for a chiller condenser pump with the VFD.Each of the models are requiring a dual feed panel to achieve the 99.999% reliability. Transfer switches and ups systems were also explored in the models, but mechanical loads should not be put on a ups system.Automatic transfer switches have too low of a reliability to be used in this application.
Thank you for your response,
Sheldon

 

[rbalex]

I am assuming you created the MTBF/MTTR SKM system model for the dual fed panel and VFD since they aren't part of the "native" SKM package. I won't bother with how you determined what they were - it doesn't matter.

Since your generator control scheme already has a stored energy component (the batteries) you probably don't need 99.999% to the charger. You need to evaluate the reliability from the battery through the control panel.

Without a fully redundant chiller package including redundant VFDs with independant sources of supply, the second model is most likely a lost cause since the chiller package itself probably isn't 99.999% reliable.

 

[barclayd]

I gave you the easiest model with the generator tie gear.The charger is for the batteries for the generator tie gear switching.

The second model is for a chiller condenser pump with the VFD.Each of the models are requiring a dual feed panel to achieve the 99.999% reliability. Transfer switches and ups systems were also explored in the models, but mechanical loads should not be put on a ups system.Automatic transfer switches have too low of a reliability to be used in this application.
Thank you for your response,
Sheldon

The batteries are usually the weak link, not so much the charger.

"A Chiller Condenser Pump" ?? One Pump??
The whole system will Never get 5-9's with one pump.
N+1 = 2 pumps - each one can carry the load.
2N +1 = 3 pumps - work on one, and still have redundancy.

By the way, how's the water supply? In a widespread power outage, your domestic water supply might not be available.

If a system requires the high reliability, that automatically requires at least daily inspections.

db

 

[sberger3]

Thanks for all the responses.
Robert: you are correct, we are interested in MTTF/MTBF/MTTR etc. for the a.c. power to the charger. Our client is only interested in the a.c. power to the charger because the rest of the generator's control power system has been analized.

For the second model, our client is also only interested in the power to the VFDs at this point in the project. SKM has a VFD in it's library and we were able to achieve 99.999% to the VFD by using a model with a dual fed distribution panel. Just as an example:we thought of using an 800 amp panel, feed it with (2) 400 amp feeds but never load the panel more than 400amps. By not loading the panel to more than 400amps, either feed could carry the full load of the panel.We can meet 240.92 for the maximum 6 disconnect rule. Any reason why this won't work?
Thank you all in advance,
Sheldon

 

[rbalex]

The 800A panel is a single-point failure source. Failure of the main bus or any line-side circuit breaker directly connected to the bus results in a total system failure with respect to the chiller. No matter how much you tweak the upstream, the chiller system cannot have a reliability factor better than the chiller itself. If a redundant chiller isn't in the picture, 99.999% reliability isn't either.

 

[Ibarra]

Sheldon,

Could you let us know why such a high availability is required?

We have worked in projects where a high availaiblity is required (hospitals, banks, etc.) but never heard something like this.

Ric

 

[sberger3]

Sheldon,

Could you let us know why such a high availability is required?

We have worked in projects where a high availaiblity is required (hospitals, banks, etc.) but never heard something like this.

Ric

I am sorry, I cannot reveal the nature of the project.
Thanks for your response,
Sheldon

 

[techntrek]

You state in post #6 that you found that "mechanical loads" shouldn't be put on a UPS. What reasoning/research led you to that conclusion? A UPS is no different than any other power source (but usually cleaner), just size it correctly for the startup surge. Inverters/UPSs are used all the time with motors...

 

[drbond24]

I am sorry, I cannot reveal the nature of the project.
Thanks for your response,
Sheldon

That means he's working for the DoD in some form. :wink: Shhhhhh.

I once managed the installation of a fully redundant UPS system big enough to back up an entire building. It was either 1 MVA or 2 MVA, I don't remember if 2 was the total or if it was just one side. In any case, it was big. They had an existing UPS that backed up the building, but it had seen the end of its expected life cycle. They bought a fully redundant system (so there were two complete systems, whether they were 1 or 2 MVA each I still don't remember) and kept their existing system as an emergency backup for the two new ones. They had redundant generators for the building as well, so the UPS system was just to carry the building until the generators wound up and came online. They told me that in the event of a total power failure from the POCO, they didn't even want to see the lights flicker inside the building. This should go without saying, but it was very expensive.

Oh, and that was for the DoD. :wink: Shhhhhh.

 

[broadgage]

As regards the battery charger, rather than any form of duplicated or changover supply to the charger, I would consider two chargers, wired to seperate power supplies.
That also protects against the battery charger itself failing.

For the pump, I would consider that any type of changover system introduces an additional point of failure.
Two supplies in paralel is also unwise IMHO.
If the two suplies are genuinely independant, then how can you be certain that they will allways be in phase ?
If the two supplies are ultimatly from the same source, then you have simply moved the common point of failure upstream a bit.

I would go for multiple pumps, one duty and one standby as a minimum, one duty and two standby would be better.
As others point out, the chiller is probably the weakest link. I very much doubt that 99.999% availability would be achieved even with a duty/standby arrangement. One duty/two standby might well be required.
These multiple pumps and chillers should be on different feeders, and preferably on diferent transformers.

Such installations require very carefull planing to ensure that any defective part may be isolated for safe replacement, without shutting down working parts.
There is little point in providing two pumps for reliabilty, and then mounting the starters in the same cabinet, and the wires in the same wireway.
How then can one starter or set of wires be easily changed out whilst the other remains live?

 

[sberger3]

You state in post #6 that you found that "mechanical loads" shouldn't be put on a UPS. What reasoning/research led you to that conclusion? A UPS is no different than any other power source (but usually cleaner), just size it correctly for the startup surge. Inverters/UPSs are used all the time with motors...

You are correct that motors could be put on ups,but the motor on the ups system would corrupt the 'clean' power and could create havock with sensitive computer equipment that normally reside on ups systems. A dedicated ups for motor loads and a separate ups for computer loads could work.

 

[sberger3]

As regards the battery charger, rather than any form of duplicated or changover supply to the charger, I would consider two chargers, wired to seperate power supplies.
That also protects against the battery charger itself failing.

There are (2) chargers for one string of batteries, goal is 99.999% for each of the (2) chargers in the system.

For the pump, I would consider that any type of changover system introduces an additional point of failure.

Two supplies in paralel is also unwise IMHO.

I see the two supply arrangement quite frequently in my field. One example is a double ended substation: each with a transformer, main breaker, connected in the middle with a tie breaker. When the tie is closed with both transformers connected together, sharing the load bus it is in 'network mode'. The substations are installed in phase with each other, designed to operate in this fashion, with AIC considerations etc. Once the tie is closed and the two transformers are connected together, they tend to regulate each other. The electrical grid in the United States is another example with multiple power stations connected to each other on the power grid.

 

[techntrek]

You are correct that motors could be put on ups,but the motor on the ups system would corrupt the 'clean' power and could create havock with sensitive computer equipment that normally reside on ups systems. A dedicated ups for motor loads and a separate ups for computer loads could work.

True of any power source, he was implying that there was a special reason why motors and UPSs specifically didn't mix. Unless I was reading too much into that?

 

[dbuckley]

For this level of reliability you need (at least) two fully independent electrical systems with no points of commonality. You will need dual UPSs in there, suitably rated, and multiple generators not paralleled. Everything independent.

Your "loads", whatever they are, need to be able to run off either or both supplies transparently and without interruption in service. So for battery chargers you need more than one charger, though it beats the heck out of me why if you have batteries you need that level of upstream reliability, I'll just take it on trust that for some reason you do. Chillers? Nothing with a motor singular will get anywhere near five nines, so you need multiple chiller thingies. Each motor needs to be able to pick which supply it wants.

The thing that kills five nines is chains of stuff. They need to be simple and clean, and end to end. So a circuit breaker that could be replaced with an isolator is a potential source of problem that should be eleiminated. You dont need a circuit breaker on both ends of a cable, just at one end. Distrust everything.

And mythbusters time: computers dont need clean power, they'll accept any old junk you've got. Theres a whole industry or ten that makes loads of money out of a problem that was solved decades ago...

 

[Ibarra]

dbuck,

Thanks for your explanation. You have conveyed in a very simple and concise way a subject that can be quite rough. Availability, reliability, resilience, redundancy, etc. are concepts that are easily confused. We have seen projects where millions have been spent in power protection and backup systems and a bad crimping shuts everything down...

In regards to PCs sensitivity, I do agree. Engineers on this field should always refer to ITIC and CBEMA curves.

Ric