When you open the door, for (variable) amounts of time, with (variable) mass of cold contents preserving the low temperature, while (variable) amounts of "hot" air, at (variable) humidity ....
Then all bets are off. I will probably have similar 4-degree rises (subsequently recovered) when I similarly poke around in my Fridge on a hot day, before preparing dinner.
But I changed to a longer control period when I saw that the system shuts down PID/Relay output power while doing analysis for about 3 seconds. If your Control Period is only 20 seconds, then you are losing a considerable portion (15%) of your maximum input energy over some multi-period amount of time. (My corresponding energy loss is only 1.5%, just 1/10 as much.) Although I haven't attempted to quantify my "recovery" from an extended open door period, it might be quite a bit faster than yours, because I sustain my input power level better.
Being more "nice" to the heater wires, and eliminating a bunch of small clicking noises from the Automotive Coil relay (DC side), are secondary benefits. (In the recommended implementation, the DC coil is switched together with the AC SSR. So it creates those clicking noises, even when you're running 120VAC on shore power.) But my #1 reason was increasing total power over time (== energy input), by reducing the frequency of those 3 second "dead" periods.
- - -
Since you are (probably) using a different mode of "settings" than I am, our software
will be different. My "software mode" drives Fridge temp down to a lower-than-expected temp (near the bottom of the range, rather than the set "desired" value). I'll SWAG that it doesn't do much of any trend analysis. And since our "heater" doesn't have a great deal of overshoot, a more simple ON/OFF method still works fine.
Perhaps it turns on the Fridge, and then slightly overshoots the "desired" temp via heater "overshoot" (lacking PID logic). Or, more likely, it stays ON past the desired set point and goes a bit further towards the minimum value, based on a programming
assumption that heat flows in at a pretty high rate, like an open grocery store freezer case.
- - - - -
If you have your temp probe in a place other than recommended (about one inch above the bottom, 1/4 to 1/3 away from the LEFT wall as you're looking inside) then you will see bigger swings than I do, even if the PID is working the same. (This location is the coldest and most stable spot in the fridge, the one where things will accidentally freeze first. That's why it is the best measurement location.)
Now, with the long explanations done
If your PID does support my setting mode, and your PT100 probe is located correctly, then you'll want to switch to my settings. A 0.3F range of door-closed temperature range
IS better than a 2.0F range, and more continuous power input
IS better than allowing a bunch of extra 'dead' periods to occur while the Fridge is too warm, and should be running.
UPDATE: PT10.0 mode doesn't work when the Fridge temp exceeds about 47 degrees, so I switched to PT100. Except for "door open events", it still holds somewhere in the 34F degree range -- but the PID doesn't show exactly where, because the display switches from 1/10 to full degrees when you change to PT100. (It just sits at 34F ... I no longer get the 34.6 versus 34.9 type of readout.)