A Design and Implementation of a Combustion Controller in Microcomputer-based Multiloop Controller MMC-90

Copyright © IFAC Control of Power Plants and Power Systems, Munich, Germany, 1992

A DESIGN AND IMPLEMENTATION OF A COMBUSTION CONTROLLER IN MICROCOMPUTER-BASED MULTILOOP CONTROLLER MMC-90 J, Cretnik, J.

Petrov~i~ and A. Bitenc

Departmellt of Computer Automatioll aNi Coll1rol, 1. StefallillStitute, Ljwbljana, Slovellia

Abstract. The J. Stefan Institute in Ljubljana has developed a microcomput er-b ased mllltiloop controller MMC-90, which is, on the one hand , aimed at scientific research work, i.e. the study of the most modern techniques of control and , on the other hand , aimed at solving actual industrial problems. This paper presents an original CACE approach used for the design and impl ementation of a combustion controller with MMC-90. The control structure was first designed and simulated using the CACSD program package ANA. Then, it was implemented in the MMC-90 and tested in a real-time hardware-in-the-loop mode. This control simulation is described and shown for the case of a medium-size industrial boiler (20 MW). The combustion was simulated on a PC and the combustion control was done by the MMC-90. The de veloped combustion control concept was also tested and later applied in the real plant. The improved combustion control on the steam boiler PK400 (PK401) at the chemical works Cinkarna Celje implemented with MMC-90 led to a 4 % fuel (heavy oil) savings . I(evwords .

Process models; process control; combustion; simulation ; control system design ;

microcomputer-based control ; control applications ; PID control.

It monitors measured values and signals alarms, performs

INTRODUCTION

open- loop and close-loop control for six control loops, regardless of whether they are cascade loops or not.

We decided to design our own microcomputer controller because in-depth knowledge of hardware as well as

So that we can work as effectively as possible with the

software of the computer system is necessary to solve more demanding control problems. Thus we have at our disposal

MMC-90 controller, we developed a CACE (computer-

a completely open system which we can , if necessary ,

aided control engineering) environment. CACE means

supplement with new control and other mathematical

computer support in the implementa tion of specifi c control

functions.

solutions in all work phases composing this process ; th ese are (Sitenc and co-workers, 1989):

Microcomputer-based multiloop controller MMC-90 can

• design of control schemes,

perform either simple or advanced control tasks in the pro-

• simulation and testing of cant rol schemes,

cess (primary chemical) and other industries , es pecially in

• impl ementation of control solutions on th e intended

small production units. It is best suited for more complex

computer and

cases where the process cannot be controlled by several

• installation of such a prepared computer at the facility.

individual controllers, yet the use of sophisticated generalpurpose process-<:ontrol computers is not justified. Typical

This articl e describes first the CACE environment for de-

for such problems are close interdependence of individual

sign and implementation of control algorithms , followed

control loops, nontrivial control algorithms and advanced

by the description of the CACE approach for design and

control structures. MMC-90 can be freely con figured and

implementation of the combustion controller. It continues

it is capable of more then just close-loop control. It works

with the realization of the control of combustion in steam

as a computer for mathematical, time and logic functions.

123

boiler PK400 (PK401)

r-- --.---

in Cinkarna Celje with our

--~ ~ ~~- --=.- ~--' HAR[)WAR E CON I ROLU~ R

technique.

~ - -j /

[11~il ~F

CACE ENVIRONMENT The main concept of the CACE approach is to use the

I

possibilities provided by today's modern computer equipment, specially personal computers (PCs). In this way, the

---u

==-

I

DATA ACOUISITION AND CONTROL ADAPTER

,; '"

prototype of a new controller can be tested rapidly, efficiently, economicaly and safely with PCs in laboratory conditions. Thus, errors which occurred during the develop-

...=.

ment stage could be eliminated before the control system

y

'E

I ~ U

t

is installed in the plant, without any danger of damaging or destroying the plant (Pfannstiel and Cretnik, 1991).

IF'AC92

Fig. 2.

The fundamental basis of the CACE approach is simulation. For this purpose, a mathematical model is needed. In most cases it is enough to have models of just parts of the

PERSONAL COMPUTER

Control of the simulated process on the PC by the HC.

_ _ .. - .- - ---

,..-'- - ' - - --- _.__ . .

plant. The control algorithms can be tested on the PC,

_._ ---

- - -----

HI>RDWARE CONTROLL ER

and in the case when data-acquisition and control adapter is connected to the PC, the developed hardware controller

r Ihl ~ - v

(HC) can be tested directly on the PC. By simulation of the processes on a PC, the same conditions for testing different control algorithms can always be used. This means that the achieved results can be easily compared. In our case we have used:

I

1. PC simulation o/the plant and the controUer:

11

this case, both the controller and the plant are simulated

...=.

on the PC . So newly developed control algorithms can be

y

tested and modified quickly and easily .

"

7

OATA ACQUIS!TtON AND CONTROL ADAPTER

LI r

The block-scheme of this possibility is shown in Fig. I. In

u ~--

~

I

~u

REAL PLANT IF"AC92

PERSONAL COMPUTER

~ --V ~

I

Fig. 3.

"b:;, t

~

----v

YE~

Control of the real plant with the HC.

Y

=-

algorithms is installed in the real plant (Fig. 3). Thanks 1.0

t

previous use of simulation (described in 1 and 2), the tests and verification of developed IIC and its control algorithms in the real plant are very short and efficient.

Fig. 1.

Simulation of the controller and the plant on a PC.

The described approach was the basic idea of the modelbased design and implementation of a modern combustion controller. It will be explained in more detail in this article.

2. Control o/the simulated process on the PC by the HC: In most cases, the control algorithm must be implemented as an autonomous control system. Therefore, the developed control algorithm must be implemented as a software

COMBUSTION PROCESS MODEL

part in HC. Consequently , the implemented control algorithm as well as the HC itself can be tested in the labora-

Due to many variable factors, occurring ill the industrial-

tory by connecting the HC with the PC (Fig. 2). Thus, the

scale furnaces, the fuel/air ratio is sometimes very difficult

errors which occur during the developmental stage can be eliminated before the new HC is installed in the plant.

to control. Either too much or too little air leads to a waste of energy, environmental pollution, and may even

9. Control o/the real plant: Finally, the developed HC with implemented control

create hazardous situations. The above problem can be successfully solved by automatic control concepts, based

124

on modern on-line flue-gas analyzers, advanced control

x(t)

= Al u(t) x(t)

algori thms and microcomputer-based hardware.

= x(t-T d)

y(t)

A2(t) v(t) x(t)

+ 82(t)

8 1 u(t) For the control design purposes, first a mathematical

+

+ (4)

v(t) ,

.

(5)

model of a combustion process was developed (Cretnik and where:

co-workers, 1985). It was adapted for a medium-size industrial boiler on which the combustion controller was


valve position, %

later installed. Its short description follows.


damper angle position, air flow , m 3 /s

u

0

There are two equal steam boilers PK400 and PK401 in

v

fuel flow, kg/s

Cinkarna Celje. They produce steam which is used in the

x,y

oxygen concentration in flue-gas, vol.%

production of titanium dioxide (Ti02). The boiler has two

Td

process time delay, s

rotational burners for heavy oil. If needed, both or just one

a

valve constant, kg/s/%

of the burners (also boilers) is in operation. The steam boi-

b- I

damper time constant, s

ler produces between 7.5 and 30 tones of steam per hour.

.1

damper nonlinear function

The steam has a temperature of between 240 and 260 0 C

AI,BI

constant parameters of combustion

and pressure of 17 bars. The air for combustion is supplied

A2,B2

time variant parameters of combustion.

by the forced-draught fan, which has a 75 kW motor. The draught of gases is natural, trough a common chimney. The exit temperature of the flue gas is around 180 oC. The

model, whose nonlinearity is given by eqn. 3. The combu-

steam boilers are an East German products, made in 1973.

stion process is described by eqn. 4 and its transport time

Eqn. 1 presents the valve model and eqn. 2 the damper

The measurement-control equipment based on pneumatics

by eqn. 5. The time delay also includes the sensors time

was also largely from East Germany. Due to wear and tear

delay.

it was partly substituted with gauges giving electrical output signals, electric transmitters, electropneumatic positi-

COMBUSTION PROCESS MODEL

oners and controllers which are realized in the MMC-90. The inputs of the model are the angle position of the fuel valve (fuel flow) and angle position of the damper. Its output is the oxygen concentration in flue-gas. The combustion process model includes three submodels (Fig. 4): the combustion model, the damper model and the valve model. The combustion model is based on stoichiometric equations supplemented with some practical phenomena which actually take place in the furnace. The combustion

Fig. 4.

model describes real combustion, which means that com-

The block diagram of the continuous model of the combustion process.

bustion is incomplete when the air ratio is low. For this reason, the parameters of combustion are rather complicated and time-dependent. The combustion model is a

The continuous combustion model was put into a discrete

first-order system in bilinear form with a time delay. The

form (sample time 1 s) and used for control design purpos-

path of fl ue-gas through the furnace and also its oxygen

es as well as in the simulation loop.

concentration measurement cause a considerable and variable time delay of the combustion process. In simulation, however, a constant time delay was used. The damper

MODEL VALIDATION

model is a first-order system with a nonlinear dependence between the damper position and air flow. The valve

As already mentioned , the developed model was adapted

model is a simple static model with linear dependence

to describe the steam boiler PK401 (PK100) at the chemi-

between valve position and fuel flow. It also gives norma-

cal works Cinkarna Celje. This model was also validated

lised fire-rate of the boiler (3.

by measurements on the mentioned real plant. The behavior of the model is shown in Fig. 5. The validation has

The continuolls model of the combustion process is

shown good dinamical behavior of the model.

described by the following equations: v(t)

= a 'Py(t) ,

(I)

~~(t)

= -b 'P~(t)

(2)

u(t)

=.1('P~(t)),

+ b 'PD(t) ,

DESIGN AND SIMULATION ON PC Considerable time delay, potential hazardous situations,

(3)

fuel

125

cost,

moderate

and

random

variations

in

the

DAMPER POSmON [%J

-

with variable gain) can trim the damper to the angle position according to the fire rate. The designed combustion control is also based on the continuous oxygen measurement of the f1ue-gas. If an error in

600

100

1100

1600

2100

2600

oxygen occurs, the damper is trimmed by the 02-control-

VALVE POSmON [%J

-

"~I

36

ler as well. The

r-"\

600

-

1100

I 2100

1600

crf('ctive (Fig . 7) . Beside t.he conventional collli>u'l.ion

v

I 2600

combustion control was developed (Cretnik, 1990). The designed combustion controller has to be implemented

13~_ \.

,

1''''

\

9

"

'"

'\

\

'""..."

-

in the real process. Therefore the structure of the combu-

...

I

stion controll er is simple, roboust and reliable. The desi-

.

\_

gned combustion controller (Fig. 7) allows:

7+-____+-____+-____+-____+-____+-____ 100

600

1100

1600

2100

• manual control of the air/fuel ratio ,

2600

• automatic control of the air/fuel ratio ,

llME [SI

Fig. 5.

("()II-

trol , realized in MMC-90 on PK400 , also special adaptive

----- 02 -MODEL [vol.%]

02 [vol.%J

.

11

is PI-controller with variab-

other modifications, which make the 02-controller more

/\ /

V

28

20 100

O~ontroller

le gain (feed-forward gain scheduling: {J and 02) and some

Comparison of the combustion model response

• manual O 2 correction of the automatic control of t he air/fuel ratio and

with the real measurement on PK401.

• automatic 02 correction of the automatic control of the air /fuel ratio.

combustion process are the major sources of problems of this control design. For this reason the control structure

COMBUSTION CONTROLLER REALIZATION IN MfJC-90

was first designed and simulated using our own CACSD program package ANA (Sega and co-workers, 1985). The developed combustion controller consists of three parts (Fig. 6): • 02-reference generator, • 02-controller (feed-back controller) and • compensator (feed-forward controller). COMBUSTION CONTROLLER

IFACt

Fig. 7.

Block diagram of the combustion controller implemented in the MMC-90.

REAL-TIME SIMULATION Fig. 6.

The block diagram of the combustion controller.

The designed combustion control structure was implemented in the microcomputer-based multiloop controller MMC-90. The desired control structure (blocks) is first

While the oxygen reference is fire-rate dependent , a refe-

edited on the PC in text or in graphic form and is then

rence generator is included in the control loop to give a

automatically converted into MMC-90 code (Bitenc and

fire-rate dependence on the set point value. Reference is

co-workers, 1989). Thi s code can then be loaded into

given in 11 discrete points with linear interpolation be-

MMC-90. The desired control structure is thus saved in

tween them. This makes for a very easy and precise refe-

~IMC-90,

rence-value profile.

changed any time via the MMC-90 keyboard.

but the parameters of the control blocks can be

The compensator as well has 11 discrete points with linear

The designed

interpolation between them. Thus the damper nonlinear

MMC-90 was then tested in a real-time hardware-in-

profile can be easily set in the compensator. In this way

the-loop mode. This real-time combustion control simula-

the compensator (feed-forward controller, P controller

tion is shown in Fig. 8.

126

combustion

controller implemented on

STEAM FLOW [I/h)

-

::~_

...

~_ r~

::~ o

300

600

900

02 [vol.%)

1200

1500

REFERENCE [vol.%]

12~

10

I

B

6

2

4--~

o Fig. 8.

300

The real-time combustion control simulation

600

900

1200

1500

CO2 [vol.%]

with the microcomputer-based multiloop

~~

controller MMC-90 and the combustion model realized in CACSD package ANA on PC.

6

I

:~ o 300

EXPERIMENTATIONS IN THE PLANT

-

The developed combustion control algorithm implemented

0.09\ 0.06

experimentally installed almost without modification in

0.03

the industrial real plant, i.e. on the steam boiler PK401 at

900

1200

1500

CO [vol.%]

_r-___-'./~2

on MMC-90 and verified with real-time simulation was

600

_ _ _ _ _ _ _ _ _ _ __

0.00 "!:---:~I--~I:__--f-I- - - - - - 4 1 - - - + 1 o 300 600 900 1200 1500

the chemical works Cinkarna Celje. For experimental pur-

TIME [sI

poses additional P /1 and I/P converters were used, while the present measurement--control equipment on the steam

Fig. 9.

boiler based on pneumatic. In addition to the described

The experimental combustion control on the steam boiler PK401 at Cinkarna Celje:

combustion controller, also the load controller was imple-

I - previous control,

mented in MMC-90.

2 - control with MMC-90.

The analyses have shown that the improved control of combustion has led to a 4

% fuel (heavy oil) savings (Fig. The complete control scheme with the MMC-90 is shown

9). Beside the direct economical effect of this develop-

in Fig. 10.

ment , the corresponding ecological aspect is also worth mentioning.

The reliability of operation and the safety of the con trolled system are two of the most important requirements. which must be satisfied by the computer control of industriill

IMPLEMENTATION IN THE REAL PLANT

plants. The MMC-90 controller is designed so that safety of plant operation is of primary concern. This requirement

The achieved results have encouraged the management of

can be fulfilled by controller warning diagnostic, signalisa-

the chemical works Cinkarna Celje to decide to put the

tion and automatic switch over of all control

whole control of the steam boiler in order with MMC-90.

100j'l5

to ma-

nual mode in case of the power or controller failure. In this way the operator, during the period of the above lllention-

The developed combustion controller currently constitutes

ed shutdown, manually controls the process through the

a part of the Combustion control system on the steam boi-

front panel of the controller. which is battery/acculllulator

ler Ph:400. It si composed of microcomputer-based Illulti-

supplied. This allows additional security in process con-

loop controller I\IMC-90 and oxygen analyzer MK200

trol. \Ve do not interfere with original protection system of

based on Zr02 cell (also developed at J. Stefan Institute).

the steam boiler because the old system of protection has

The Combustion control system (MMC-90) on PK400

worked well and thus can remain unchanged in the cttrrent

performs the following control tasks:

process control. • boiler load control. • air/fuel ratio control with additional 02 control , • drum level control.

CONCLUSION

• fuel pressure control and • fuel temperature control.

The following tasks were performed, using the approach

127

presented in this article (Bitenc and co-workers, 1989;

phase of simulation and implementation) and, of course, with the process itself and its model.

Pfannstiel and Cretnik, 1991): • setting up objectives and requirements for control

The presented model-based approach, which belongs to

functions,

the

• process analysis, mathematical modeling and derivation of a simplified model,

field

of

Computer-Aided

Control

Engineering

• synthesis of control structures and algorithms,

(CACE), allows us to design a modern combustio~ control for any boiler quickly, economically and safely. The appro-

• testing of control algorithms by simulation and their modification,

ful in the design and implementation of the controller in

• real-time synchronization of control tasks,

any other processes (specially chemical and process industries) as well.

ach used and the environment developed can be very help-

• transformation of control algorithms into controller code, • integration with other functions on the controller and

REFERENCES

• final testing, debugging and fine tuning.

Bitenc, A., S. Strmcnik, J. PetrovCic and J. Cernetic (1989). A computer-aided engineering approach to design and implementation of advanced control algorithms, Proc. of Signals & Systems, 3, pp . 25-37, . Brighton. Cretnik, J. , S. Strmcnik and B. Zupancic (1985). A model for combustion of fuel in the boiler. Proc. of 3 th Symposium Simulationstechnik, pp. 469-473, Bad . Munster a. St . Ebernburg. Cretnik, J. (1990). A case study of self-tuning control of combustion process . Prep. of MIM-S2'90, pp. VIII.S.5. H5, Brussels. Pfannstiel , D. and J. Cretnik (1991). In circuit simulation of industrial processes with personal computer, Proc. . of 6 th MELECON, vol 11., pp. 876-879, Ljublja.na. Sega, M., S. Strmcnik, R. Karba and D. Matko (1985) . Interactive program package ANA for system analysis and control design. Prep. of CADCE'85, pp. 145-150, Copenhagen.

The iterative nature of the above tasks and the element of real-time demand an effective computer support in this field. We have developed our own computer-supported working environment for implementation of advanced control structures. This environment is composed of hardware (a development computer (PC) and microcomputer-based multiloop controller MMC-90 (HC)) and corresponding software tools (CACSD program package ANA, automatic code generator for MMC-90 controller and interactive graphical editor for constructing control schemes BLOK). Such an environment has the possibility to communicate with the control engineer (in the phase of development and simulation of the control scheme), the operator (in the

~~~~'~------------------~

REC\.Jl.ACVSf(.I. SHE"" PARNECA 1<0ll.A PK400 V C'1NfUrRNI eruE POONtA. l a.OK! V£eVoHCH£CA WlKRQR,A,tlnW.Nl~[CA R£CU\..ATORJA w"'-to

.. =- .

r:.~cc,.owc l . ll.1H
Fig. 10.

Realization of the control structure for steam boiler PK400 in the form of M MC-90 functional blocks.

128