Research PID control. That allows your setpoint to match the measured value.
Once you know what 0 and 100% are and have a way to measure temperature your almost set,
bang bang control is the simplest, but if you have any thermal mass, overshoot is possible.
You have something called "slow cycle triac". This will essentially make n zero crossings 100%
You have phase angle firing (PAF) where at some point you turn on the triac. Your also working with 1/2 cycles.
You also end up with a table of V^2 vs firing angle vs 0-100%. Power is proportional to V^2 and the resistance of the heating element is constant.
here are two links:
https://ccipower.com/product-category/zero-cross
https://ccipower.com/product-category/phase-angle
I don;t think the PI has the interrupt latency to do either of these.
In the real industrial world, you come up with a lot of choices. PAF is difficult to use with transformers. You can start at how to control and optionally measure power with a 0-100% control signal. A 0-100% control signal on V^2 with a current meter and manual (knob) current limit is one way to go. Tungsten loads or an incadesent lamp dimmer are ususally cheap because 0-100% is not anywhere close to 0-100% power.
I'm guessing that a 0-100% that's time proportional should work for you where the minimum resolution is 1/2 cycle, but you may not be able to support that either so maybe 1sec on, is the minimum on/off time. So, maybe 100 s is a good place to start for 0-100%.
You probably won;t get or use an interrupt on a zero cross, but that's ideal. 0-100% would basically be related closely to 0-100% power. If you had better interrupt latency, then you could use a 1/2 cycle.