The MPC will disregard the heating of the standard thermal model because it is not sensible, and will assign the constant 0.05 heating. however, the doing_nothing of the standard model is consistent and will hence be used . However, now heating and doing nothing are inconsistent.
For now, i will not use a thermal model if it is not consistent or sensible for any of the actions.
Might want to consider other possibilities. such as taking the maximum temperature increase across all actions and assign it to heating, etc.
Currently, this example is a problem:
Standard Thermal Model: cooling = -0.2 doing nothing = 0.1 heating = 50
Average Model: cooling = -0.2 doing nothing = 0.00001 heating= nan
constant model: cooling= - 0.05 doing nothing = 0 heating = 0.05
The MPC will disregard the heating of the standard thermal model because it is not sensible, and will assign the constant 0.05 heating. however, the doing_nothing of the standard model is consistent and will hence be used . However, now heating and doing nothing are inconsistent.