See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/278704907

Simulation of MIT Rule-Based Adaptive Controller of a Power Plant

Superheater

Chapter · January 2012

DOI: 10.1007/978-3-642-27552-4_65

CITATIONS READS

6 2,007

3 authors:

Zdenek Machacek Martin Pieš

VŠB-Technical University of Ostrava VŠB-Technical University of Ostrava
55 PUBLICATIONS 395 CITATIONS 60 PUBLICATIONS 332 CITATIONS

SEE PROFILE SEE PROFILE

Stepan Ozana
VŠB-Technical University of Ostrava
115 PUBLICATIONS 586 CITATIONS

SEE PROFILE

All content following this page was uploaded by Stepan Ozana on 23 December 2015.

The user has requested enhancement of the downloaded file.

 Simulation of MIT Rule-based Adaptive Controller of a
Power Plant Superheater

Zdenek Machacek1, Martin Pies1, Stepan Ozana1

1
VSB-Technical University of Ostrava, 17. listopadu 15
70833 Ostrava, Czech Republic
{zdenek.machacek, martin.pies, stepan.ozana}@vsb.cz

Abstract. MIT rule is one of the basic techniques of adaptive control. It can be
embedded into a general scheme of circuit with MRAC structure (reference
adaptive controller). This paper deals with simulation of MRAC by use of MIT
rule for a superheater which is a crucial part of a coal-fired power plant. A
superheater has been chosen as a typical practical real plant with time-variant
parameters and thus suitable to be controlled by adaptive control mechanism.
The paper gives a comparison of adaptive control with MRAC and original PID
control currently implemented in a chosen output superheater in Detmarovice
power plant. Matlab&Simulink environment is used for a simulation of the
control circuit described in this paper.

Keywords: Adaptive control, MRAC, Superheater, MIT rule,

MATLAB&Simulink

1 Introduction

Adaptive control systems with reference model have been originally designed for
control of continuous (deterministic) system with unknown dynamics or possibly
variable dynamics. Specification of the problematic is based on a reference model
which represents ideal trend line of control circuit as a response to reference signal ur.
General block scheme is introduced in Fig. 1.
Control system contains two loops: inner and outer one. The inner loop is standard
feedback from controlled process. The outer loop contains tuning mechanism that
allows tuning of controller parameter In order to keep the control error e between
process value y and reference model output minimal. The outer loop thus also covers
a controller. The key problem is to design tuning mechanism in order to keep the
control loop stable and force control error to zero value which is not trivial problem.
The next part of the paper will be devoted to the overview of adaptive algorithms and
their application for control the superheater which has been chosen as a typical
practical real plant.
Fig. 1. General scheme of adaptive control circuit with reference model

2 Adaptive Control Concepts

A regulated plant is treated as an adaptive system, which means as a system that can
modify structure or parameters of its components (mostly controllers) according
change of regulated process in order to keep the original quality of control. Adaptive
systems can be divided into several categories regarding a given goal (stabilizing
systems, invariable systems, invariant systems, optimal adaptive systems).

There are several often used methods and tools for design of adaptive control:
• Methods of continuous identification (for example least square method), or
other methods for state-space approximation
• Matlab&Simulink tools, particularly Simulink Design Optimization (iterative
algorithms based on Hess matrix)
• Predictive controllers based on minimization of objective function over a
prediction horizon
• Other specific methods (gradient approach, approach based on stability
theory)

Since this paper extends previously published work under the common project
devoting modeling and simulation of the dynamics of superheaters and heat
exchangers [1], [2], [3], [4], [5], a superheater was also chosen as a typical system
with variable parameters and dynamics. Adaptive character of this plant is mostly
caused by changes of heat transfer coefficients that affect the dynamics of the system
significantly. The mathematic background that gives a description of the superheater
is beyond the scope of the paper, it is given in detail in [3], [4].
3 Control Circuit with a Superheater

A rough outline of the control circuit of output superheater of Detmarovice [8] power
plant is shown Fig. 2. Its basic task is to keep constant steam temperature T1(L,t) =
540 °C at the superheater outlet (corresponding to set-point Twms = 540 °C), while hot
flue gas Tfg ≐ 1100 °C is brought to the superheater, heating up the working media.
The temperature regulation is carried out by injection of a certain amount of the water
in the mixer, corresponding to manipulated value in the control circuit, using a
cascade control (branched-off circuit) structure.

Fig. 2. Control circuit scheme for output superheater.

Control circuit of the superheater includes two basic control loops. The main
control loop keeps the constant value of temperature T1(L,t) at the output of the
superheater. The controller of this loop is nonlinear PID controller with the
approximate derivative. Parameters of this controller vary with the steam flow Mmix
through the superheater, and they are assigned by function generators. In a wide range
of operational steam flow through the superheater these parameters of PID are
constants, therefore it was replaced by linear PID controller with multiplicative syntax
of control law algorithm.

4 Gradient Approach of Adaptive Control for Superheater

This approach is one of the basic but efficient methods of adaptive control. It is
introduced in [6],[7] and its tuning mechanism is known and referred to as MIT rule,
named after Massachusetts Institute of Technology:
d e (1)
  e
dt 

Components of the vector  e  represent sensitivity derivatives of control error

with respect to adjusted parameter θ. Approximations of sensitivity derivatives can be
generated as the outputs of linear system driven by inputs and outputs of controlled
process. Parameter γ determines the speed of adaptation. MIT rule can be interpreted
in the following way: Let’s suppose that parameters θ change much slower than other
variables in the system. Sometimes it might be suitable to tune the parameters
according direction of negative gradient e2 to ensure the small value of control error
square. Tuning mechanism is actually composed of linear filter generating sensitivity
derivatives, product block (signal multiplicator and integrator. Parameters are then
introduced to the control law by the second multiplicator. This mechanism is the core
of many adaptive systems with reference model. Fig. 3 represents the block scheme of
this structure. Tuning mechanism represents adaptation of direct feed-through gain
based on gradient approach with the use of MIT rules. It can be derived as follows:

Fig. 3. Error model, the core of tuning algorithm in gradient approach.

Let’s denote transfer function of the model as  0G  s  , process is described by its

transfer function G(s), see Fig. 4. Then control error is computed as:


e  y  ym  L1 G  s  U r  G  s   0U r 
 L1 G  s U r    0      0  L1 G  s U r  


 G  s  U r 
 
0
  
0
    0 L1   L1 GmU r  

  0

  0 (2)
  0   0
 L1 Ym   ym
 0
0
e 1 d  eym
 y , 
  0 m dt 0

Fig. 4. Block scheme of MRAC based on MIT rule.

5 Setting up Reference Model

Reference model expresses the shape of the output response y that is considered to be
ideal response to reference signal ur. In many cases including this paper it specifies
the ideal step response of closed control circuit, and it is mostly described by the
transfer function. Determination of this transfer function is one of the crucial parts of
design of adaptive control scheme in this paper. The “PID” marked line in Fig. 6
shows step response of branched-off control circuit shown in Fig. 2 as reaction to 1°C
step change of set-point Twms (from 540 to 541 °C, input shifted to zero), ouput is also
shifted and therefore denoted with ∆ sign. The shape and character of the trend line is
caused by nonlinearities of the components and also by two control loops (faster
and slower). Ideal reference model to be substituted to the adaptive control scheme
can be setup as the approximation of this trend line by a suitable transfer function.
This can be obtained by a specific identification method or by ident tool in
Matlab&Simulink environment. The detailed description of this identification is
beyond the scope of the paper. However, using ident tool the approximation was find
in the form of the 1st order transfer function, which was declared as reference model
to be substituted into the block scheme in Fig.4 and simulated:

G s 
1 (3)
172.14s  1

6 Simulation of the Adaptive Control Circuit

The control circuit with MRAC according Fig. 1 has been implemented in Simulink.
The trend lines in Fig. 5 and Fig. 6 have been shifted down by constant value 540, just
for clarity reasons (therefore denoted with ∆ sign).

Fig. 5. Step response of closed-loop circuit with MRAC.

Fig. 5 represents a closed-loop step MRAC response corresponding to 1°C step
change of set-point Twms (from 540 to 541 °C, input shifted to zero) for several values
of γ adaptation parameters. Fig. 6 then shows a comparison between current original
PID control and MRAC structure, for γ = -5∙10-8.

Fig. 6. Step response of real closed-loop circuit compared to MRAC.

7 Conclusion

The paper presented a new approach using MIT rule as one of the main adaptive
control approach based on MRAC structure. Simulation experiments work with the
real input data and parameters from output superheater in Detmarovice power plant.
They show that presented approached could be working for a given technology.
Control system includes one parameter γ which stands for adaptation speed. Its value
is crucial to produce acceptable and usable trend lines of process value, and in this
case it was found out manually. Once it is tuned, the step response of closed-loop with
MRAC structure gives quite good and fast results with almost none overshoot, see Fig.
6. The steady state is reached in less than 600 seconds which is important from
technological point of view. However, tuning γ still leads to a compromise, as it can
be seen in Fig. 5, because obtaining a small overshoot is redeemed by rising
undershoot for different values of γ. This parameter can be also tuned by adding some
other techniques to the control scheme, for example fuzzy-MRAC modification. From
the point of view of integral quadratic criteria the results of PID and adaptive control
are almost the same, the important fact is that adaptive control structure can treat
varying parameters of superheater unlike PID control.

Acknowledgment. The work was supported by the grant “Simulation of heat

exchangers with the high temperature working media and application of models for
optimal control of heat exchanger”, No.102/09/1003, of the Czech Science
Foundation.
 References

1. Nevriva, P., Ozana, S., Pies, M., Vilimec, L. Dynamical Model of a Power Plant
Superheater. In WSEAS Transactions on Systems 9 (7), pp. 774-783. Issue 7, Volume 9,
2010, available from WWW: http://www.wseas.us/e-library/transactions/systems/2010/89-
878.pdf. ISSN: 1109-2777.
2. Nevriva, P., Ozana, S., Pies, M. Simulation of Power Plant Superheater Using Advanced
Simulink Capabilities. In INTERNATIONAL JOURNAL of CIRCUITS, SYSTEMS and
SIGNAL PROCESSING, pp. 86-93. Issue 1, Volume 5, 2011, dostupný z WWW:
http://www.naun.org/journals/circuitssystemssignal/19-545.pdf. ISSN: 1998-4464.
3. NEVRIVA, Pavel; OZANA, Stepan; VILIMEC, Ladislav. Verification of the Dynamical
Model of a Power Plant Superheater. In KALLEL, Ali, et al. ADVANCES in
DYNAMICAL SYSYSTEMS and CONTROL : 6th WSEAS International Conference on
DYNAMICAL SYSTEMS and CONTROL (CONTROL'10). [Kantaoui,Sousse,Tunisia] :
WSEAS Press, 2010. s. 54-59. ISBN 978-960-474-185-4, ISSN 1790-5117.
4. Ozana, S., Pies, M. Modeling and Simulation of Power Plant Superheater in Simulink. In
Proceedings of the International Conference on Circuits, Systems, Signals (CSS'2010).
Sliema, Malta. p. 180 - 183, September 15th - 17th, 2010. ISBN 978-960-474-226-4, ISSN
1792-4324
5. PIEŠ, Martin, Ozana, S. Simulation of Power Plant Superheater by Simulink S-functions. In
15th IEEE International Conference on Emerging Technologies and Factory Automation
(ETFA'2010). Bilbao, Spain : Escuela Técnica Superior de Ingeniería de Bilbao. 1x CD-
ROM, September 13rd - 16th, 2010. ISBN 978-142446850-8.
6. ASTROM, K. J. , WITTENMARK,B. Adaptive Control, Addison-Wesley, 1989, 2d ed.
1994.
7. ŠIMANDL, M. Adaptivní systémy. ZČU Plzeň, 1993. ISBN 80-7082-098-5
8. CEZ GROUP. THE DETMAROVICE POWER STATION. [cit. 2011-08-03],
<http://www.cez.cz/en/powerplants-and-environment/coal-fired-
powerplants/cr/detmarovice.html>, 2011.

View publication stats