Measuring UPS Efficiency – Help!


Written on April 17, 2010 – 1:41 pm | by ac45

One aspect of determining our Power Use Effectiveness (PUE) is to understand how much energy is “wasted” by our UPSes.  This is especially important as we look to move from many small rack-mounted UPSes to a more central approach.  We’ve been quite confused in reading about the relationships of Real and Apparent power and Power Factor and how that relates to efficiency.

In Power Basics for IT Professionals (HP, October 2007)  this description of Power Factor is provided (my highlighting):

The power factor (PF) of a device is a number between zero and one that represents the ratio between the real power in watts and the apparent power in VA. A power supply that has a PF of 1.0 indicates that the voltage and current peak together (the voltage and current sine waves are always the same polarity), which means that the VA and watt values are the same. A device with a Power Factor of 0.5 would have a watt value that is half the VA value; for example, a 400VA device with a Power Factor of 0.5 would be a 200W device.

A common misconception is that the power factor and the power supply efficiency are related, but this is not the true. Power supply efficiency is the ratio of output power in watts to input power in watts at peak efficiency. For example, a typical white box power supply with a peak efficiency of 75 percent would waste at least 25 percent of the incoming energy by converting it to heat that must then be dissipated. HP ProLiant server power supplies all have peak efficiencies of 85 percent or greater, which increases the amount of power that performs useful work.

Devices with a low power factor, on the other hand, do not waste energy. Unused energy is simply returned to the utility and is not paid for by the customer. Utilities charge for true power used as measured in kWhours, not in VA. The main costs associated with a low power factor are for higher amperage circuits to deliver the same amount of true power as a device with a power factor closer to one.

Power supplies for servers usually contain circuitry to correct the power factor (that is, to bring input current and voltage into phase). Power-factor correction allows the input current to continuously flow, reduces the peak input current, and reduces energy loss in the power supply, thus improving its operational efficiency.  Power-factor-corrected (PFC) power supplies have a power factor near unity (~1), which allows smaller circuits to be used. Using energy-efficient PFC devices, including UPSs, can lead to significant cost savings for data centers where the incoming feeds are measured in megawatts. As a standard feature, power supplies for ProLiant servers all contain circuitry to correct the power factor (that is, to bring input current and voltage into phase).

The above statement about apparent power (VA) vs. real power (W) and how it’s billed for is somewhat contradicted in Horowitz & Hill, The Art of Electronics, 2nd Edition (ISBN 0-521-37095-7) page 35 (again, my highlighting):

Power factor is a serious matter in large-scale electrical power distribution, because reactive currents don’t result in useful power being delivered to the load, but cost the power company plenty in terms of I^2R heating in the resistance of generators, transformers and wiring.  Although residential users are billed for “real” power [Re(VI)*], the power company charges industrial users according to the power factor.  This explains the capacitor yards you see behind large factories, built to cancel the inductive reactance of industrial machinery (i.e., motors).

Ok, so we now understand that Power Factor matters and that we should make sure our power systems provide Power Factor Correct (PFC) if possible.   We are already measuring Power Factor with our Wattnode meters as well as with the SNMP-monitored UPSes.

An article in Wikipedia (a truly unbiased source of peer-reviewed information clearly written by a vendor of IGBT technology:-) states:

Double Conversion Efficiency Savings

The major discussion point of an ECO UPS is its efficiency savings. A true “double conversion” ECO friendly UPS must use input and output IGBT technology, only then can a UPS hope to reach a 95% AC to AC efficiency, this efficiency is typically more than 10 to 20% higher than a six or twelve pulse alternative. An ECO IGBT UPS should also benefit from a high input power factor (.99pf) and low reflected harmonics, these two crucial features, apart from reducing running costs, also dramatically lower installation costs as smaller cables and MCB’s (Fuses) can be used.

An ECO UPS should always have a IGBT input stage and as a benefit of this, reflected electrical noise back to the utility supply should be less than three percent (<3%), this low reflected distortion results in electronic equipment up stream from the UPS being much less susceptible to interference (ideal for hospitals).

Another important consideration, with regard to THD is that most European countries are looking at imposing penalties on companies that excessively pollute the electrical grid, distortion levels should be low from any reputable ECO UPS and as such should not fall foul to these proposed financial penalties.

A secondary but very important feature of a quality ECO UPS is when considering using a generator to support your UPS, because reflected distortion is so low a substantially smaller engine and alternator can be used. Typically when calculating the size of a generator to run with a none ECO UPS a factor of 2.5 would be applied, example, a standard 100kVA UPS would require a 250kVA generator, the same 100kVA requirement but with ECO UPS features would only require a 120kVA set, a factor of 1.2, this is 52% smaller and offers profound ecological and financial benefits.

So we can conclude that:

  1. Approaching a unity power factor is ideal mainly due to the capital cost savings in conductor sizes needed and in the resistive heating losses wasted in the power distribution network.  We still want to be good for the planet, even if the utility bills us for W rather than VA — but I expect they bill for VA.  And,
  2. Power Factor has little to do with UPS conversion efficiency, which is somewhat described in the Wikipedia article.

How to calculate UPS efficiency given the data our UPS monitoring equipment provides?

We still have more work to do to measure and understand our UPS conversion losses but should probably be looking at them in turns of Input vs. Output Apparent Power as a most conservative case.

Here’s a screen shot from the web management console of our new Eaton blade UPS which is still very lightly loaded:

BladeUPS

I need some help here! How is the Apparent Power calculated?  Given this is a 3-phase system, if it were balanced, and we knew the overall Amperage, the VA would be V x A x 1.732 (√3).   But we don’t have just A, but rather A1, A2 and A3 (and similarly V1, V2, and V3) for the three phases.  Furthermore, it is pretty unbalanced with a range between 31.8 A on one leg and of 16.6 A at the other extreme.  Working backwards from what the UPS tells us is the Apparent Power of the Output, 11515/210/1.732 = 31.66 A which is close to the worst case.  Using the worst-case approach for the Input Apparent Power (and given the currents are pretty close to one another), let’s say the Input Apparent power is 209 * 21.3 * 1.732 = 7710 VA.  Which is less than the output Apparent Power, so clearly I’m doing something wrong here.

Below is  a screen shot from the older Eaton 9315 which powers the mainframe, p-Series and EMC storage systems (we just recently installed a network monitoring card in this 5-10 year old device):

eaton-9315

These numbers seem to make more sense.  Calculate an Input Apparent Power of say 207*78*1.732 = 27,965 VA.  Output Apparent Power is shown as 23,900 VA, which when working backwards gets 23,900 VA/208 V/1.732 = 66.3 A!  Maybe the amperages have to be averaged: (76+70+53)/3 = 66.3.   In any case, we can now guess that our UPS conversion efficiency is 23,900/11,515 = 85% which seems to be in the expected ballpark.

One thing to note is that both Eaton’s (claim) to present a balanced load to the utility even though they are hiding a pretty unbalanced load (+- 15 A and 23 A for the two UPSes, respectively).  This should result in a desirable near-zero neutral current presented to the upstream distribution network.

Quoting from that HP paper (with my use of italics to highlight a controversial statement):

The power factor (PF) of a device is a number between zero and one that represents the ratio between the real power in watts and the apparent power in VA. A power supply that has a PF of 1.0 indicates that the voltage and current peak together (the voltage and current sine waves are always the same polarity), which means that the VA and watt values are the same. A device with a Power Factor of 0.5 would have a watt value that is half the VA value; for example, a 400VA device with a Power Factor of 0.5 would be a 200W device.

A common misconception is that the power factor and the power supply efficiency are related, but this is not the true. Power supply efficiency is the ratio of output power in watts to input power in watts at peak efficiency. For example, a typical white box power supply with a peak efficiency of 75 percent would waste at least 25 percent of the incoming energy by converting it to heat that must then be dissipated. HP ProLiant server power supplies all have peak efficiencies of 85 percent or greater, which increases the amount of power that performs useful work.

Devices with a low power factor, on the other hand, do not waste energy. Unused energy is simply returned to the utility and is not paid for by the customer. Utilities charge for true power used as measured in kWhours, not in VA. The main costs associated with a low power factor are for higher amperage circuits to deliver the same amount of true power as a device with a power factor closer to one.

Power supplies for servers usually contain circuitry to correct the power factor (that is, to bring input current and voltage into phase). Power-factor correction allows the input current to continuously flow, reduces the peak input current, and reduces energy loss in the power supply, thus improving its operational efficiency.  Power-factor-corrected (PFC) power supplies have a power factor near unity (~1), which allows smaller circuits to be used. Using energy-efficient PFC devices, including UPSs, can lead to significant cost savings for data centers where the incoming feeds are measured in megawatts. As a standard feature, power supplies for ProLiant servers all contain circuitry to correct the power factor (that is, to bring input current and voltage into phase).

———-

The above statement about apparent power (VA) vs. real power (W) and how it’s billed for is somewhat contradicted in Horowitz & Hill, The Art of Electronics, 2nd Edition (ISBN 0-521-37095-7) page 35:

Power factor is a serious matter in large-scale electrical power distribution, because reactive currents don’t result in useful power being delivered to the load, but cost the power company plenty in terms of I^2R heating in the resistance of generators, transformers and wiring.  Although residential users are billed for “real” power [Re(VI)*], the power company charges industrial users according to the power factor.  This explains the capacitor yards you see behind large factories, built to cancel the inductive reactance of industrial machinery (i.e., motors).

——————–

From Wikipedia (an truly unbiased source of peer-reviewed information clearly written by a vendor of IGBT technology:-)
at http://en.wikipedia.org/wiki/Eco_UPS#Double_Conversion_Efficiency_Savings

Double Conversion Efficiency Savings

The major discussion point of an ECO UPS is its efficiency savings. A true “double conversion” ECO friendly UPS must use input and output IGBT technology, only then can a UPS hope to reach a 95% AC to AC efficiency, this efficiency is typically more than 10 to 20% higher than a six or twelve pulse alternative. An ECO IGBT UPS should also benefit from a high input power factor (.99pf) and low reflected harmonics, these two crucial features, apart from reducing running costs, also dramatically lower installation costs as smaller cables and MCB’s (Fuses) can be used.

An ECO UPS should always have a IGBT input stage and as a benefit of this, reflected electrical noise back to the utility supply should be less than three percent (<3%), this low reflected distortion results in electronic equipment up stream from the UPS being much less susceptible to interference (ideal for hospitals).

Another important consideration, with regard to THD is that most European countries are looking at imposing penalties on companies that excessively pollute the electrical grid, distortion levels should be low from any reputable ECO UPS and as such should not fall foul to these proposed financial penalties.

A secondary but very important feature of a quality ECO UPS is when considering using a generator to support your UPS, because reflected distortion is so low a substantially smaller engine and alternator can be used. Typically when calculating the size of a generator to run with a none ECO UPS a factor of 2.5 would be applied, example, a standard 100kVA UPS would require a 250kVA generator, the same 100kVA requirement but with ECO UPS features would only require a 120kVA set, a factor of 1.2, this is 52% smaller and offers profound ecological and financial benefits.

—————-

So we can conclude that:

1. Approaching a unity power factor is ideal mainly due to the capital cost savings in conductor sizes needed and in the resistive heating losses wasted in the power distribution network.  We still want to be good for the planet, even if the utility bills us for W rather than VA — but I expect they bill for VA.
2. Power Factor has little to do with UPS conversion efficiency, which is somewhat described in the Wikipedia article.

On 4/17/2010 11:52 AM, Alan Crosswell wrote:

See page 14 of the HP article "Power Basics for IT Professionals" linked
from the bottom of


		
		

https://wiki.cc.columbia.edu/greendc:start#literature_review