You are on page 1of 4

CHAPTER 8 183

ADVANCED REGULATORY CONTROL


The term advanced control is often used, abused, misused, and overused. Even among knowl-
edgeable control engineers, the term does not have a consistent meaning. It is used to refer to
everything from cascade control, which can be implemented with analog instrumentation, to
optimization and model-based predictive control, which usually require a host computer inter-
faced to a lower-level distributed control or data acquisition system.
We use the term advanced regulatory control. By this we refer to a collection of control tech-
niques, from ratio and cascade up through decoupling control and dead-time compensation.
The unifying concept underlying the techniques that we include in advanced regulatory control
is that they can all be implemented with hardware modules or, what is more likely, with soft-
ware function blocks in a microprocessor-based control system. With technology such as
FOUNDATION Fieldbus, it is also possible for these techniques to be implemented with
function blocks distributed in fieldbus devices. We will not focus on the platform for imple-
mentation, however, other than to say that the techniques covered here are on a level below the
control tasks that are traditionally reserved for a host computer, such as optimization, schedul-
ing, and model predictive control.
The topics that we consider to be in the category of advanced regulatory control include:
Ratio control,
Cascade control,
Feedforward control,
Override control,
Control of multiple input, multiple output processes (decoupling control),
Dead-time compensation and elementary model-based control.
The last topic in this list will serve as a lead-in to model predictive control, for which an over-
view will be presented in chapter 15. Model predictive control can truly be called advanced
process control rather than advanced regulatory control.
8
Wade04.book Page 183 Thursday, April 15, 2004 12:20 PM
BASIC AND ADVANCED REGULATORY CONTROL: SYSTEM DESIGN AND APPLICATION
184 CHAPTER 8
These topics can all be considered as tools in a control system engineers toolkit. Just as a good
workman is competent in the use of his tools, so the control engineer should know how and
when to use each of these tools. But, like the workman, he or she should not feel obligated to
use each tool on every project.
Before we begin the discussion of individual topics, we will first investigate the motivation for
using advanced regulatory control (or any form of advanced control, for that matter). Suppose
that at some time (to be designated as time 1) we observe the record or trend display for a
process variable, along with the controller output signal, shown in Figure 8-1. Both variables
are steady, and the process variable is at set point.
At some time later (time 2), we again observe the record or trend of these variables. Again,
both variables are steady, and the process variable is right on set point. But between time 1 and
time 2, a significant load change must have occurred because the controller output at time 2 is
considerably different from its value at time 1.
Let us suppose that we do not know what happened between time 1 and time 2, and we have no
historical data that can be recalled (not a very realistic situation, but go along with me any-
way). Do you think it is possible for the controller to have been tuned well enough for the pro-
cess variable to have remained on set point between time 1 and time 2, that is, to have ridden
through the load change without deviation?
Regardless of your intuitive answer, lets do a bit of formal analysis. Using Equation 4-3, we
compute the controller output at time 1 as follows
1
:
Figure 8-1. Two Separate Observations of the Behavior of a Control Loop
1. We can use a proportional+integral controller here, rather than PID. The derivative mode would have no role in
the discussion that follows.
Process
Variable
Controller
Output
Set
Point
m
1
m
2
$m
1 2
TME
?
Wade04.book Page 184 Thursday, April 15, 2004 12:20 PM
BASIC AND ADVANCED REGULATORY CONTROL: SYSTEM DESIGN AND APPLICATION
CHAPTER 8 185
(8-1)
Note that we are talking about a particular time (time 1), so we use a definite integral here,
with subscripts and limits on the integral sign, which represent a particular time. The lower
limit of integration, time 0, can refer to the last time that the controller was switched into
automatic, and the term m
0
refers to its initial output value at that time.
Since the PV and SP were equal at time 1, the e
1
is zero. Equation 8-1 can be simplified as fol-
lows:
(8-2)
Similarly, since the error at time 2 is also zero, the controller output at time 2 is given by:
(8-3)
This integral can be broken into two regions, giving:
(8-4)
We are interested in the change in controller output, which is given by:
.
Subtracting Equation 8-3 from Equation 8-4 yields:
(8-5)
Equation 8-5 highlights the fact that in order to have feedback control action, there must be an
error. In this case, the error is represented by the area under the curve between times 1 and
times 2. There are an infinite number of possible trajectories of the process variable from time
1 to time 2. One possible trajectory is shown in Figure 8-2. All possible trajectories would
1
1 C 1 0
I
0
1
m K e edt m
T

= + +

1
C
1 0
I
0
K
m edt m
T
= +

2
C
2 0
I
0
K
m edt m
T
= +

1 2
C C
2 0
I I
0 1
K K
m edt edt m
T T
= + +

2 1
m m m =

2
C
I
1
K
m e dt
T
=

Wade04.book Page 185 Thursday, April 15, 2004 12:20 PM


BASIC AND ADVANCED REGULATORY CONTROL: SYSTEM DESIGN AND APPLICATION
186 CHAPTER 8
have the same net area, given the required valve movement from m
1
to m
2
and the tuning
parameters K
C
and T
I
.
We can surmise (and this is confirmed by practical experience) that if the load change is very
gradual, the deviation between set point and measurement would be very small. Our analysis is
consistent with this. If times 1 and 2 are widely separated, the same net area under the curve is
still required, but since it is distributed over a long time period, the maximum deviation is
reduced. (Think of cutting a rubber sheet the shape of the shaded area in Figure 8-2, then
stretching the ends of the cut-out sheet.) Thus, for a slow disturbance, feedback control by
itself is adequate. For a faster disturbance, something more is needed.
This discussion points out the curse of feedback controlto have control action, a penalty
must be paid in the form of an error (deviation) in the feedback loop. This fact presents both a
challenge and an opportunity for advanced regulatory control. How can you obtain the
required control action (valve movement) without paying the feedback penalty?
This challenge is the theme of the following chapters.
Figure 8-2. The Feedback Penalty
Process
Variable
Controller
Output
Set
Point
m
1
m
2
$m
1 2
TME
FEEDBACK
PENALTY
Wade04.book Page 186 Thursday, April 15, 2004 12:20 PM

You might also like