Do you have a BrewPi Spark? What kind of heater do you have?
Can you zoom in on a few peaks?
Please give some more info to help us help you.
Spark with the photon. Heater is a 70W 70x30cm pad. I’ll pop up some more pics when I get home. Thanks 
The fridge responds pretty slowly to the heater, so I think what is happening is that the integrator slowly winds up, only to slowly decrease later and causing some overshoot.
Can you see at which intensity the heater is running at the peak intensity? You can see this is you refresh the device manager with read values, or in the control algorithm tab:
heater1–>outputActuator–>value
Please also take note of p, i and d of heater1. I suspect your proportional part § is too low and increasing Kp will help.
Also the new beta release might help, because it has some measures to reduce overshoot.
Please make a note of the state when it crosses through zero, not on a positive peak.
Ah, now I see what is happening.
When you are on target, the input error is still 1.2 degrees. This is because the PID input is filtered, which creates a small delay. If you reduce the filtering on the cooler and heater PID input, it will not lag as much and will stop heating sooner.
I think 9 or 18 seconds is the right filter delay for you.
Thank you! I don’t have a cooler installed by the way. Tomorrow I’ll have more spare time and will put the beta on. For now I have lowered by filter delay settings for the heater. Thank you again☺
The new beta is much better, thank you again!
Good to hear. Do you have an updated chart?
I am still not happy with that result. You should be able to get a flat line.
Did you change the input filtering on the heater1 pid?
Can you post the heater1 part of the JSON again?
Oops I forgot to reply to this, it’s flattened out a lot over the Easter weekend. Will try and remember to post pic tonight.
Are you running the new beta?
Please also answer my previous questions:
I am surprised your fridge responds to heating so slowly. Do you have the fridge sensor in air?
The response time is about 30 minutes. So your integrator time constant of 10 minutes is too fast. It will cause the integrator to try to correct a steady state error which is not really a steady state error, just a slow response.
So increase Ti to 3600, so it will increase and decrease a lot slower (an hour). Or even 2 hours.
I would set Td to 600. Set the derivative filtering to 139 seconds. With a low setting of 9/18 seconds, you will mostly see bit flips as short bumps in the derivative instead of actual rises.
That’s more like it