Advanced Controllers

[OllieC]

We have to control HCl, SO2, NOx, NH3 among others
in flue gases. The controllers are all independent PI loops basically.
The disturbances are very large, sometimes unmeasured.
The APC system is highly non-linear with delays in
measurement and control.

Surely some sort of MIMO adaptive control or model
predictive control would be better? Does anyone have
any experience of this? I've only heard of its use in
chemical and pharma applications.

Ollie

 

[CARF]

put your question on measurement & control forum?

Anyway,

1st advice: if you can actually MEASURE the disturbances, you can improve control dramatically by adding FEEDFORWARD control to your traditional PI-feedbackcontrol loops.

2nd) General recommendation: first investigate all traditional control strategies (PID-tuning, feedforwards, gain schedulers, cascade control) before going to "model" based control. This control method is only as good as your model. Developping an accurate model for your specific application takes years. => My experience: a pain in the @^%$. Operators will go to manual as soon as you're in the taxi.

Good luck,
MVD

 

[OllieC]

Thanks MVD

We are using feedforward with the gain adjusted by a
feedback loop currently, but the feedforward does not
take account of delays (which are a similar Tc to the
process changes). A bit of modelling suggested
moving the instruments and using cascade control would
be better.

Anyway, I take your comments re. MPC. I can't see our
DCS supporting it without some major rewiring.

cheers again
Ollie

 

[CARF]

Hi,

Yes just moving the sensors closer to your actuator can decrease transportation lag times and improve your control loop (need new tuning!).

This is a very good site:

http://www.controlstation.com

Hope to read your results some time,
MVD