[Nut-upsdev] Megatec and Batteries
Arjen de Korte
nut+devel at de-korte.org
Mon Oct 27 22:54:14 UTC 2008
Citeren Zeljko Baralic <zeljko.baralic at pakom.com>:
>> You can't calculate battery charge of a
>> UPS by looking at the battery voltage alone
>>
>> It physically just isn't possible.
> But we may get some very close values with accuracy of maybe 90%.
Read the above remark again. Not by looking at the battery voltage
alone. The internal resistance in the battery and the dependency of
voltage on the load applied (to name a few) prevent you from doing
that.
>> Although in principle it would be possible to get a reasonably
>> accurate figure if there would be no current going into or out of the
>> battery (also known as the open cell voltage), this is impractical in
>> a real life UPS application. You can't just switch off the load, wait
>> for a while to redistribute the chemicals inside the batteries and then
>> measure the open cell voltage.
> You can never get actual health state of battery by measuring just
> open cell voltage because some bad batteries can suffer from voltage
> memorizing effect. Known battery load must be attached to it when
> measuring its voltage.
Read carefully, I wasn't writing about the health of the battery.
[...]
> When UPS operates on battery, this energy is mostly used for
> Inverter (part that creates OUT when there is no IN power to UPS)
> than for supplying internal logic of UPS and also some goes to
> thermal loses of conversion...
In order to measure battery charge, knowing how much Watt hours it has
delivered doesn't help much. We want to know the remaining charge
(which is load independent), not the amount of power we already
extracted (which is load dependent). Having said that, the efficiency
of the energy conversion is unknown anyway (depending on size of the
UPS) and so is the charge the battery can hold. Note that NUT won't
remember state/values between restarts, so anything that isn't stored
in the UPS, will be lost forever.
> What is good about megatec is that you can always get Load% and you
> know rated power/voltage out of UPS.
Not at all. Quite a number of UPS'es will happily report a load of 0%
(or any other fixed number) at all times. The protocol may be able to
deal with this, but many implementations don't.
> This means (simple example): UPS load 50% on 220V and max out power
> is 1000VA you have 1000VA * 0.5 = 500VA. You can say that this is
> energy from batteries.
No you can't. You're confusing apparent and real power here. The load
the UPS is reporting is invariably apparent power, where the batteries
will only have to supply real power.
> If bat bank is 24V nominated and you have read with megatec 24.4V
> than divide 500VA with 24.4 you will get 20.5A current from
> batteries. You will have more accurate value if 1000VA *
> PF of 0,6 = 600W than * Load% 0.50 = 300W than add 10% on internal
> looses you will get cca 330W divided by 24.4V gives some current
> 13.55A from batteries. Since there is a big difference I will check
> in my lab which is more accurate and give a post.
The power factor is dependent on the load and this isn't reported, so
this value isn't available for most ordinary users. Therefor, I
restate that it is impossible to calculate the battery current from
the values provided by the megatec protocol and hence coulomb counting
is not possible.
> Anyway Arjen, from my point of view creators of megatec protocol
> where tried to make something in between contact-clousure and
> inteligent signaling and I think that they succeded in this. They
> give some minimal info about a state od device for disaster
> protection.
Absolutely. I'm not trying to put down the makers. In most cases,
knowing the battery charge is irrelevant, since NUT will only use the
low battery warning signal for shutting down. But in some cases, where
the remaining runtime on battery after signaling low battery is
insufficient for an orderly shutdown of the connected systems for
instance, this is not enough. In that case, a ballpark figure isn't
good enough.
--
Please keep list traffic on the list
More information about the Nut-upsdev
mailing list