I started this experiment on 6-28-2018. Its been almost five years now and the house has yet to burn down as a result. I had some free time and decided to make some measurements a couple of months early.
...
I may have missed it, but what do you use for balancing the pack/array? Also, how much are you de-rating the capacitors? Have you tried higher loads on the UPS?
Sorry for the year delay but I don't check this thread often.
I did not care about efficiency and balanced them by measuring the leakage when they were new and increased it using a resistor in parallel with each. When I measured the voltage drop today they are all still well balanced.
I had measured the UPS charge and dropout voltage to determine the capacitance required to achieve the hold up time. It looks like I doubled the capacitance required.
I have ran it at higher loads but continue to collect the data a 300W to give some idea how the parts are degrading. I wasn't expecting the hold time to be as good as it is and had to repeat the test. Looks like about 11 seconds. Consider we started out with about 17 second holdup, we have lost a fair amount. Because you asked, I increased the load to a bit over 700W and the holdup time dropped to about 3.7 seconds. I let it recharge the bank and tried it again with roughly the same result.
As we still have these short dropouts, I plan to let it continue to run.