You mean 300W (over 8 to 12 hours daylight, 2400 to 3600 Wh), and 1kW (over 16 or 12 hours night, 16kWh or 12kWh)?
A zero-export grid-tie system size for 300W would be pretty small. A couple panels, and there are some cheap inverters. Hardly worth the trouble. A grid-tie inverter that us UL-1741 listed and does do backfeed of grid is something you might get away with and not have utility notice.
An average of 15 kWh/day drawn from grid
This correct, the current lighting consumption is between about 15.8 kWh and 18.4 kWh a day,
however I am also upgrading to LED to lower the consumption. To give you some perspective,
the billing rate is based by Tiers of 8 kWh a day, so the consuption is already about two Tiers,
and the rate for each Tier increases with usage, similarly any drop of consumption can be noticeable.
The lighting system is independent from the other utilities, in partcular from the service plugs.
(Well except one plug that was used to vacuum one of the hallways, and I had to fix this.)
So the Solar system can be easilly keept
off grid and this simpler than to have to deal with any Grid-Tied meters and subscriptions.
Note: The Solar system can be used as an emergency backup, so I wonder how a GT is designed in case of an outage?
I imagine that Smart meters must be designed to avoid any Solar production getting sent back to the grid in this situation.
About $0.35/kWh for nighttime draw, $5/day $150/month $1800/year. OK, $200 actual.
The alternate rate schedule with low price at night may save you money, but lighting between 4:00 and 9:00 PM (peak time) and 3:00 to 4:00 & 9:00 to Midnight (one of the schedules) will cost more. GT PV would help in the afternoon until sun gets low, especially with panels aimed SW.
If there is any surplus available, well a GT could have some advantages, but surplus could be used locally, like heating hot water.
I might then consider adding some heat pumps, since the natural gas used for getting the hot water is also based on Tiers....
A 48V 14kWh DIY LiFePO4 battery might cost around $2000 or so. If it does last 10 years, 3500 cycles, then you have saved a significant amount compared to utility rates.
Life tests of commercial batteries had only 5% of them reach expected cycle life without failure. I'd say don't count on DIY LiFePO4 lasting more than 25% of anticipated life (so 2.5 years for $0.20/kWh) which may be slightly better than break even. It may still do 5 to 10 years, in which case you come out ahead. So long as failures are BMS and cells are undamaged, you can DIY repair and are more likely to get the payback.
I am currently evaluating various 48 V batteries, mostly based on 280 Ah (or 310 Ah), so using 3.2 V Cell, I should get
about 14.3 kWh (or 15.9 kWh), but I plan to discharge from 100% to 20%, about 225 Ah (or 248 Ah),
or from 3.65 V to (80 % of 3.65 V - 2.5 V cut off) = 2.73 V, to avoid too much degradation.
About ambient temperature, in San Francisco it is not be very common to reach 50 F or 10 C, especially inside a building.
Using a 80% discharge, there should be about 11.5 kWh (or 12.7 kWh) available. This might be a little bit short in winter,
since the night consumption would increase to reach a total of 18.4 kWh a day, but I cannot see too much any other battery
combination unless I have to double the capacity. However a larger battery capacity would allow getting
a depth of discharge (DoD) lesser than 0.6 C which could improve the lifespan of the battery cells.
Load is (1 kW / 48 V) = 21 A, so for a 280 Ah battery, the discharge rate of 0.07 C is well below the typical 1 C discharging.
About charging, since full capacity is (3.2 V x 16) = 51.20 Volts x 280 Ah = 14.336 Wh, and considering a 20 % of various conversion
and transport losses, a Solar production of 17.2 kWh would be required which is possible even in winter
when the expected production from the 18 PV system is still
19 kWh. But a full charge of 100% would be also much slower,
but sufficient to fully charge the battery, especially if the battery was only discharged to 20%.