PID is not always the best algorithm to apply to temperature control.
Temperature control is extremely problematic because there are long transport lags involved in the process which need to be included in the model of the process to be controlled.

This is not the right place to discuss the control problem however suffice it to say that the heat control problem should be modelled as a first order system with dead time.

You must estimate the dead time and the thermal time constant of the process.

Dead time is time difference between application of heat and actual start of change in temperature.

Thermal time constant is rate of change of temperature from the time it starts to actually change to approximately 67% of final temperature if the heat were to be constantly applied.

Sounds very complex, highly mathematical and a pain in the back side to get going.
What is worse, even after all of the above is done, the tuning is still to be performed and then monitored to make sure the control loop behaves properly.

There is a more modern approach to the PID ( old analogue way ).

A newer method is the so called "take back half"

http://electronicdesign.com/an...-temperature-control-algorithm

http://www.edn.com/design/anal...ack-half-convergence-algorithm

It would represent a minimum problem to implement the algorithm in LDmicro.

Good luck