Solar made 57.5 kWh, 56% less than Monday, and 47% of what the house used came from solar and the battery.
No 15-minute readings from 12:15 AM to 1:15 AM, so the lines break there. The day's totals come from the meter counters.
High 97°F, average 85°F · sunlight 3.5 kWh/m² · rain 0.01 in · gusts to 12 mph · humidity 80% · 117 lightning strikes nearby.
Day totals come from the battery system's lifetime counters, read once the day closes. Until then, or when that row is missing, they're added up from the 15-minute readings instead. Self-powered is the share of house use that didn't come from the grid: 1 − grid_import_kwh ÷ home_kwh. The forecast on Produced is the one fetched before the day began, scored by the forecast model fitted without this day (Now explains the model). The dashed Expected line spreads it over the day the way the array's day usually builds up: at each quarter hour, the median share of a day's total made by then over the 9 closed days before this one.
House use by source splits each 15-minute reading: solar goes to the house first, then charges the battery, then goes to the grid; the house's remaining use comes from the battery, then the grid. The energy balance adds those splits up, so charging and discharging aren't netted against each other. It uses the 15-minute readings, so it can differ by a few kWh from the meter counters in the tiles; 2.6 kWh moved between the battery and the grid directly and isn't in either bar.
Appliance activity comes from the thermostats, pool controller through Home Assistant, polled about every 10 minutes; a bar means the appliance was running at a poll.
Weather is from the Tempest station next to the house, 144 readings this day. Sunlight is the average irradiance × 24 h (AVG(solar_radiation_w_m2) × 24 ÷ 1000).