Just to note: not just colder, but the dew point is also now moving frustratingly into the region where ice builds up (becoming the “frost point”) on the fins of the outdoor unit, which then starts acting as a (very good) insulator, and then eventually also blocks airflow……
This effect (even prior to the actual defrost which is easily spotted in the data), will drag efficiency down.
The effect is most pronounced directly prior to the defrost in the visual data, where input electrical power ramps up to drive the fan(s) and compressor harder and faster to compensate for the reduction in heat exchange and air flow rate through the back/side of the unit.
Eventually it (should) give up and say to itself (algorithmically and via ice build up sensor(s)): “I’m better defrosting (taking heat from inside the house to melt the ice formation) than carrying on at the reduced CoP.
Some (our Daikins for example as I see it it milder weather too) have algorithms that are somewhat dumb, and for example defrost when it’s warm outside and no physical frost formation.
Vinny mentions this here:
And Alison Bailes (USA) discus this issue here:
So just note this point when analysing your data feeds as we are now moving into a very different season with different observable characteristics on heat pump running.