Confirm calculation of battery size for actual energy data

[greenlinker]

Hi everybody! I have a 3.6 kW system (14 solar panels + Growatt inverter), and I am considering adding a battery to it. As in The Netherlands the new electricity meters have a debugging port that spits out every 10 seconds, I have been getting data for a couple of years of the excess energy pulled / raw pushed into the grid. This is my actual pushed/pulled measurements sampled at a period of 1h, across a year (disregard the whole in between, my measuring system went down):
pushed_pulled.png

Today I have analyzed the data to try to estimate the size of a battery for my current system, and what I have done is: at the end of each time bucket (1h), I have filled up with the excess energy a virtual battery and then I have subtracted from it what was pulled from the network originally. This produces the following data, for the pushed/pulled and the battery charge level:

simulated.pngWith this data, it seems that a 3 kWh battery should be enough for my system. Do you guys think this conclusion is sound?

[davidcheok]

Take a week to log the energy usage from the point when the sun is no longer able to cover (beginning of import) and when it can cover (export). I suggest doing it in 6 hr intervals.
Alternatively, just use the total imports divided by the number of days and add a certain percentage on top. Maybe 50%?
If from that chart, seems like peak is about 3 kWh on days where there is no production so 3.5 kWh seems like a good number. Always try to avoid completely discharging the battery.