I have this boiler controlled manually at the moment, and I am wondering what would be the most proper approach to automate this process.
The boiler in this question basically burns fuel (saw dust, wood chips or coal dust) to produce high pressure steam that is consumed by other manufacturing processes.
Fuel is transferred to the Burning Chamber by a (VFD controlled) Conveyor belt. Here it will be burnt to produce heat to make steam. The desired level of heat in the chamber to be maintained is 800-1000 degree Celsius.
The Fresh Air Fan (also VFD controlled) works in harmony with the speed of the fuel Conveyor belt to provide sufficient oxygen for the combustion. Its speed is also dictated by the level of oxygen inside the burning chamber indicated by the Oxygen Sensor.
The Exhaust Fan (also VFD controlled) speed is only dictated by the vacuum level inside the burning chamber, indicated by the Vacuum Sensor. The speed of the Exhaust Fan is regulated to maintain a fixed level of vacuum inside the burning chamber.
As steam is consumed by the consuming manufacturing activities, the steam pressure measured by the Pressure Gauge will be lower, indicating the need to speed up the Conveyor to produce more steam.
Naturally, I would be thinking of applying several separate PID control loops for each sub-process of this system. However, the difficulty of this problem lies in the HETEROGENEOUS nature of the fuel. Incoming fuel varies in moisture content and calorific value (i.e. moist fuel will drop the heat inside the chamber and requires more time to produce heat, while fuel with a low calorific value produce less heat than that with a high calorific value.)
Maybe a machine-learning approach of some kind is best suited for this problem? Are there other control approaches that are best fit to this problem?