Last semester we measured exactly this in our capstone project, rough figures are as follows:Your cycle: 3.2 s × average ~110 mA × 3.3 V ≈ 1.1 J per transmission. At one transmission every 10 minutes → average ~2 mW, plus the deep sleep current. An indoor 10 cm² panel away from the window gives 50-200 µW, so it is 10 times too little. A small 0.5 W panel with direct sunlight through a window + supercapacitor + a harvesting IC such as BQ25570 is enough even in winter. A TEG attached to the radiator gives a few mW at a 20-30 K difference, but in summer the heating is off, and as Gizem Hanım said, forget about vibration.The real gain is in the number of transmissions. After waking up, do NTP (you are connected to Wi-Fi anyway), collect the readings in RTC memory and send a single POST once an hour:

[{"ts":1793000000000,"values":{"temperature":21.4,"humidity":44}},
 {"ts":1793000600000,"values":{"temperature":21.5,"humidity":44}}]

If there is no clock, leave out ts; the send time is then stamped. Also send the supercapacitor voltage as supercap_v, and you will see from the graph which node is sufficient.