Posted by John on November 10, 2008, 11:02 am
I was using Homer (http://www.nrel.gov/homer ) to figure out why a
colleague's PV/battery installation was doing so poorly. Since the
Homer software is new to me I manually did some of the calculations
using the algorithms given in the help file.
I was suspicious because Homer was predicting a 10-year life for the
Trojan T-105 batteries when I knew they were being murdered in
significantly less than this. It seems the battery model is horribly
flawed. I can get the same result as Homer does by using Homer's model
algorithm. And I understand the basis of the model. But the result
just does not reflect reality.
Homer works on the basis of a battery being able to cycle (store &
regurgitate) a certain amount of energy before replacement. Once it
has cycled that amount then it is deamed to have met its life limit.
This life is then limited to the manufacturer's float/shelf life if
that's less. The estimate of the energy a battery can cycle is derived
from the manufacturer's cycle testing and the calculation is straight
forward enough. And from those results the idea of a fixed amount of
energy cycling is in reasonable agreement with the data. But why is
the battery life forecast just so wrong?
My suspicion is that when the maker does cycle testing the battery
sample is discharged to whatever state of charge is being tested and
then immediately recharged for the next cycle. In real life, when a
battery reaches a low state in a PV-only system it's because of crappy
weather and it may not get fully charged for days - until the sun
shines. This leaves the plates sulphating up, seriously reducing
battery life. The model takes no account of this effect at all.
1) Have I properly identified the cause of the discrepancy?
2) How can such a detailed application that's spent a decade & a half
in development still have such a shoddy battery model? After all,
PV/battery-only systems are not uncommon.
Having bagged Homer's simplistic battery model, I still commend the
Homer people for their work on the PV modelling. Their use of
insolation databases on the internet, modelling of indirect lighting
and panel orientation is impressively comprehensive and I recommend
the software. The "stable" version seems quite stable but the current
beta has temperature modelling. Pity it keeps crashing if I dwell for
long in certain menus.
Posted by Ron Rosenfeld on November 10, 2008, 12:41 pm
I have found that Drs. Lilienthal and Lambert, whose email addresses are in
the HELP/About section, have been extremely responsive in answering any
questions I've had about HOMER. I would ask them about the battery model.
With regard to the crashing, I found it when in the Battery tab of the
simulation. Here is what Tom Watson wrote me last August about that:
"The crash on the Battery tab of the Simulation Results window happens due
to one of the table cells having a tool tip that exceeds the maximum size
of 80 characters. We have just made the change necessary to avoid that
crash, so the next version of HOMER will not have that problem. Tool tips
appear when you hover over a cell on a table, so in the short term you can
avoid this problem by not hovering your cursor over the tables on the
Posted by John on November 11, 2008, 7:28 am
On Mon, 10 Nov 2008 07:41:06 -0500, Ron Rosenfeld
Yes. I contacted Tom Lambert. As you say, very helpful.
Tom didn't dispute the issue with modelling batteries that get deep
discharge and actually took up my suggestion to at least add a user
warning in this situation. Of course, adding another level of
sophistication to the model would be even better. :-)
I also found some calculation bugs on the Create New Battery page
which he's been keen to follow up. He asked me to send my .hmr file so
he could examine what was happening. I gotta laugh - I replicated the
problem using one of his own sample files!
Yes, I got the same story on the crashes. Nice to have a workaround.
Posted by Eeyore on November 12, 2008, 5:56 am
Oh that kills them in no time at all.
Never go below say 30% capacity.
Posted by Ron Rosenfeld on November 12, 2008, 11:00 am
That's good to know. I hope they can add another level of sophistication.
It can be difficult to compare HOMER results with real life when dealing
with something like battery life. HOMER does come pretty close to modeling
my own system. But, of course, I haven't had to replace batteries yet, so
can't check that parameter against reality.