[ Follow Ups ] [ Post Followup ] [ Control Consulting ] [ FAQ ]
Posted by Don Snowden on March 30, 2000 at 10:44:15:
In Reply to: Scope of design of MPC controllers posted by Gary Cera on March 29, 2000 at 07:55:32:
The key decision for designing a multivariable controller mainly boils down to economic opportunities & robustness verses complexity. Let me first say, before I go on, that the Controller Design Phase is one of the more difficult phases of the project and the more experience personnel, the better. A good control engineer will be thinking about “how is this controller going to work when I turn it on” from day one of the project! Keep in mind that in the above statement about robustness, I’m stating that the system is probably more robust (if the project is executed properly) under the larger controller verses the smaller controllers scenario.
In any case, you don't want to create an overly complex control system just to make it large. On the other hand, you don’t want to give up on economic opportunity due to a more-simplistic approach. If the unit is broken down in too many pieces (I’m assuming that we’re referring to a fairly interactive unit), then either one or two things is going to happen: one, the controller limits are going to be set conservatively or two, one of the controllers is going to be turned off since they “fight” each other. In most cases, the technology and computer power is available to handle most of the bigger problems we face today. In addition, it is possible to “present” the information to the Operators in simpler sections (e.g. Rx/Regen subcontroller and Fractionator subcontroller) for them to be better able to interact with the control system. Thus the only challenge is for the control engineer to “think and work” on a more unit-wide basis verses an individual unit operation basis.
As you pointed out one of the biggest factors is controller interaction (and constraints). If two pieces of the plant have serious interaction (e.g. the Rx/Regen and the Fractionator) then it can be difficult to break these pieces apart. In fact, it is more difficult to treat them separately then to treat them as one controller. In fact, in some cases, the Gas Plant must be included in the Rx/Regen/Fractionator Controller. For example, when the Absorber/Stripper column and the debutanizer column is reboiled by the Fractionator pumparounds AND there are constraints in the gas plant (e.g. reboiler valve constraints) then it may be necessary. You can see that if you don’t put them together then you will not be able to maximize the total profitability of the unit (e.g. you will have to back away from the true limit on the constraints). On the other hand, if there are no, or very limited, reboiler constraints on the Gas plant reboilers then it may be possible to treat the gas plant and the Rx/Regen/Fractionator as separate controllers.
Overall, the only guideline I can think of is that “You don’t want to make the controller any bigger than it needs to be, but make sure you have captured enough of the interactions & constraints to maximize the economic benefits”. That probably doesn’t help much but it you want to outline a particular example, one of us will try to outline a potential strategy.