Format And Type Fonts

CCHHEEMMIICCAALL EENNGGIINNEEEERRIINNGG TTRRAANNSSAACCTTIIOONNSS

VOL. 39, 2014

A publication of

The Italian Association

of Chemical Engineering

www.aidic.it/cet
Guest Editors: Petar Sabev Varbanov, Jiří Jaromír Klemeš, Peng Yen Liew, Jun Yow Yong

ISBN 978-88-95608-30-3; ISSN 2283-9216 DOI:10.3303/CET1439133

Please cite this article as: Kortela J., Jämsä-Jounela S., 2014, Fault tolerant model predictive control for the biopower 5

CHP plant, Chemical Engineering Transactions, 39, 793-798 DOI:10.3303/CET1439133

793

Fault Tolerant Model Predictive Control

for the BioPower 5 CHP Plant

Jukka Kortela*, Sirkka-Liisa Jämsä-Jounela

Aalto University School of Chemical Technology, P.O. Box 16100, FI-00076 Aalto

jukka.kortela@aalto.fi

The main aim of control of the BioGrate boiler is stable energy production, where a fuel bed height sensor

is a critical element in the control of the BioGrate boiler and its faulty operation should thus be avoided. A

fault tolerant model predictive control (FTMPC) has been developed to accommodate the fault in this fuel

bed height sensor by active controller reconfiguration. The proposed FTMPC is tested with the BioPower 5

CHP plant data by simulation and finally the results are presented, analysed, and discussed.

1. Introduction

In energy production, the substitution of fossil fuels by solid biomass has increased in an attempt to reduce

2
CO emissions. Great efforts have been made to develop biomass boilers and significant technological

steps have been achieved in recent years (Grölles et al., 2014). One of the newest successful processes

which can burn biomass with a moisture content as high as 65 % is a BioGrate boiler technology

developed by MW Power (Boriouchkine et al., 2012).

Due to the long time delay and the moisture in fuel feed, development of FTMPC for the BioGrate boiler

requires special considerations especially in the modelling of the BioGrate combustion. Bauer et al. (2010)

derived a simple model for the grate combustion of biomass based on two mass balances for dry fuel and

water. The model was subsequently verified by experiments in a pilot scale furnace with a horizontally

moving grate. Based on the literature (Johansson et al., 2007), they suggested that the overall effect of the

primary air flow rate on the thermal decomposition of dry fuel is multiplicative. In addition, the test results of

Bauer et al. (2010) showed that the water evaporation rate is mainly independent of the primary air flow.

Based on the models, Grölles et al. (2014) implemented a model based control strategy in a commercially

available small-scale biomass boiler. Test results showed that the model based control strategy could

always provide the required load whereas the conventional control could not avoid a feed temperature

drop of more than 7 °C. In addition, better control of the residual oxygen and the control of the air ratio led

to lower emissions and higher efficiencies. Moreover, the control was able to handle the step-wise change

of the fuel water content from 26 % to 38 % and vice versa without difficulties. As the developed control

requires the knowledge of variables like the mass of water in the water evaporation zone and the mass of

dry fuel in the thermal decomposition zone but as only the feed temperature could be measured, the

extended Kalman filter was incorporated into the model to estimate the current state of the furnace. Kortela

and Jämsä-Jounela have presented a solution for this issue in (Kortela and Jämsä-Jounela, 2012) where

these variables are estimated by using fuel and moisture soft-sensors.

A fuel bed height sensor is a critical element in the control of the BioGrate boiler and for optimal energy

production its faulty operation should thus be avoided. The effect of occuring faults, especially for

sustainable energy development, can be prevented by using fault-tolerant control (FTC) strategy. These

strategies are generally categorized into passive and active approaches as described in (Zhang and Jiang,

2008). An active FTC strategy for the Naantali refinery dearomatisation process was developed by

Sourander et al. (2009). While a novel model-free fault tolerant wind turbine control strategy was proposed

by Jain and Yamé (2013). The analysis of the results showed that the FTC strategies were capable of

handling faults in the online quality monitoring in the former whilst maintaining a desired power generation

794

in the latter in the event of faults. In addition, Kettunen et al. (2011) presented an active integrated fault-

tolerant MPC for an industrial dearomatisation process. On the basis of the economic evaluation of just

one feed grade, the annual estimated savings of the integrated FTMPC were predicted to be up to as

much as USD 143,000.

In this paper a FTMPC strategy is proposed to accommodate the fault in fuel bed height sensor by active

controller reconfiguration. The paper is organized as follows: Section 2 presents the BioPower 5 CHP plant

process. FTMPC strategy is presented in Section 3. The test results are given in Section 4 and Section 5,

followed by the conclusions in Section 6.

2. Description of the BioPower 5 CHP plant and its control strategy

In the BioPower 5 CHP plant, heat for electricity generation and a hot water network is obtained by direct

combustion of solid biomass – bark and woodchips – which is fed into the BioGrate together with

combustion air.

The essential components of the boiler are an economizer, an evaporator, a drum, and primary and

secondary superheaters. Figure 1 illustrates the boiler of the BioPower 5 CHP plant. Feed water is

pumped into the boiler from a feed water tank. The water is first run into the economizer (4), which is

heated by means of flue gases.

From the economizer, the heated feed water is fed into the drum (5) and along downcomers into the

bottom of the evaporator (6) tubes that surround the boiler. From the evaporator tubes, the heated water

and steam return back into the steam drum, where they are separated. The temperature of the steam is

increased first in the primary and the secondary superheaters (7) before the superheated high-pressure

steam (8) is led into a steam turbine, where electricity is generated.

2.1 Current control strategy of the BioPower 5 CHP plant
The main objective of the BioPower 5 CHP plant is to produce a desired amount of energy by keeping the

drum pressure constant. This is achieved by controlling boiler power by managing the stoker speed, as

well as the primary and secondary air intakes.

The fuel feed is controlled by manipulating the motor speed of the stoker screw to track the primary air flow

measurement with required amount of primary air and secondary air for diverse power levels, specified by

air curves. The set point of the secondary air controller is adjusted by the flue gas oxygen controller to

provide excess air for combustion and enable the complete combustion of fuel.

Figure 1: 1. Fuel, 2. Primary air, 3. Secondary air, 4. Economizer, 5. Drum, 6. Evaporator, 7. Superheaters,

8. Superheated steam

795

3. Fault Tolerant Model Predictive Control of the BioGrate boiler

A fuel bed height sensor is a critical element in the control of the BioGrate boiler and for optimal energy

production its faulty operation should thus be avoided. The strong coupling between the primary and

secondary air flows necessitates that the plant operates in a very narrow range of the air distribution. As a

result, the amount of fuel on the grate needs to be strictly controlled around a set point, which can be done

by using the fuel bed height sensor or by controlling the fuel bed height indirectly through the combustion

power variable.

The overall structure of the FTMPC follows an active reconfiguration-based FTC scheme, relying on

directly adjusting the controller itself by changing the controller structure though the parameter vector
p

r .

The controller reconfiguration-based FTC scheme is presented in Figure 2.

The proposed strategy of the BioGrate boiler uses controller reconfiguration-based FTC for the fuel bed

height sensor fault where active FDD is utilized. The failure of the measurement is detected by calculating

a root mean square error of prediction (RMSEP) index of the fuel bed height state values from two different

control observers and comparing this value to a detection threshold:

RMSEP

ii




 1

2,1,
ˆˆ

(1)

where n is the number of the samples in the test data set,
1,

ˆ
i

x is the estimated fuel bed height state of the

first MPC configuration, and
2,

ˆ
i

x the estimated fuel bed height state of the second MPC configuration. The

two control observers are run in parallel with an input disturbance model for fault detection purposes and

for smooth interconnection of the two controllers.

The proposed FTMPC consists of two different MPC configurations. The models of the first MPC

configuration are structured as follows: The primary air flow rate and the stoker speed ( u ) are the

manipulated variables; the moisture content in the fuel feed and the steam demand are the measured

disturbances ( d ); and the fuel bed height and the steam pressure are the controlled variables ( z ).The

models of the second MPC are configured as follows: The primary air flow rate and the stoker speed ( u )

are the manipulated variables; the moisture content in the fuel feed and the steam demand are the

measured disturbances ( d ); and the combustion power and the steam pressure are the controlled

variables ( z ).

Figure 2: FTMPC of the BioPower 5 CHP plant

796

There is a direct relation between the combustion power and the fuel bed height controlled variables.

Therefore, the fuel bed height can be controlled and the fault in the fuel bed height sensor can be detected

through the combustion power variable. However, a variation in a fuel moisture content needs to be

considered:

 
wevhthdpathdwf

mhcmcqQ  0244.0  [MJ/s] (2)

where
wf

q is the effective heat value of the dry fuel (MJ/kg),
thd

c is the thermal decomposition rate

coefficient,
pa

m is the primary air flow rate (m
3
/s),

thd
 is the coefficient for a dependence on the position

of the moving grate,
h

c is the fuel bed height coefficient, h is the fuel bed height (m), and
wev

m water

evaporation rate (kg/s).

The MPCs utilize the linear state space system (Maciejowski, 2002):

1k
x 

kkk
EdBuAx 

(3)

k
z 

kz
xC

where A is the state matrix, B is the input matrix, E is the matrix for the measured disturbances, and
z

the output matrix.

3.1 Regulator
The system of Eq(3) is formulated as:









jjk

zk
uHxACz (4)

where
jk

H


are impulse response coefficients. Therefore – using the Eq(4) the MPC optimization problem

with input – the input rate of movement, and output constraints are:

min  



N

k
SkQkk

urz
z

(5)

s.t.
1k

x 
kkk

EdBuAx  , 1,,1,0  Nk 

z 
kz

xC , Nk ,,1,0 

maxmin

uuu
k
 , 1,,1,0  Nk 

maxmin
uuu

k
 , 1,,1,0  Nk 

maxmin
zzz

k
 , Nk ,,2,1 

where  is the objective function, N is the prediction horizon,
k

r :s are the target variables,
z

Q is the

tracking error weight matrix, S is the move suppression factor weight matrix, and the
1


kkk

uuu .

4. Description of the simulation and testing environment

A simulation model of the BioPower 5 CHP plant was built in the MATLAB environment. In addition, the

code for the FTMPC was developed. Parameters of the models of the water evaporation, the thermal

decomposition of the dry fuel, and the drum were determined by using the data of the BioPower 5 CHP

plant. Moreover the plant was further modified by installing 8 pressure sensors the BioGrate to measure

the fuel bed height pressure.

5. Test results of the FTMPC strategy

In order to demonstrate the effectiveness of the proposed FTMPC strategy, the performance of the

FTMPC was evaluated using BioPower 5 CHP plant simulator in a MATLAB environment.

797

The input limits were 0
min,1

u , 4
max,1

u , 03.0
min,1

u , and 03.0
max,1

u [kg/s] for the stoker speed;

0
min,2

u , 4
max,2

u , 03.0
min,2

u , and 03.0
max,2

u [kg/s] for the primary air.

In the nominal case, the output limits were 2.0
min,1

y , 1
max,1

y [m] for the fuel bed height; and

0
min,2

y , 55
max,2

y [bar] for the drum pressure.

In the reconfiguration, the output limits were 0
min,1

y , 30
max,1

y [m] for the combustion power; and

0
min,2

y , 55
max,2

y [bar] for the drum pressure.

In the test scenario, the power demand was changed from 12 MW to 16 MW at time step 200 s. The effect

of a drift-shaped fault in the fuel bed height was tested with and without the FTMPC strategy active. An

upward drift-shaped gradually increasing fault was introduced into the fuel bed height measurement,

starting from the time step 500 s. Then, the power demand was changed from 16 MW to 12 MW during a

time period of 800 – 1,000 s. As can be seen from the Figures 3-6, without the FTMPC the drift fault had

the effect that both the primary air and the fuel bed height started to increase rapidly. With the FTMPC,

both the primary air and the fuel bed height remained within their normal operation limits, thus improving

the reliability of the control system. The fuel moisture content was changed at a time 700 s but it didn’t

affect the fuel height as it was estimated by fuel moisture soft-sensor.

Figure 3: Reactions of moisture in fuel, dry fuel flow,

and fuel bed height to drift fault in the fuel bed height

sensor without the FTMPC active

Figure 4: Reactions of pressure, combustion power,

and primary air flow to drift fault in the fuel bed

height sensor without the FTMPC active

Figure 5: Reactions of moisture in fuel, dry fuel flow,

and fuel bed height to drift fault in the fuel bed height

sensor with the FTMPC active

Figure 6: Reactions of pressure, combustion power,

and primary air flow to drift fault in the fuel bed

height sensor with the FTMPC active

Figure 7 shows RMSEP index of different fuel bed height state values of MPC 1 and MPC 2. The detection

threshold value was chosen so – to be twice as high as the fuel bed height state differences caused by

power demand changes in different MPC configurations.

798

Figure 7: RMSEP index of fuel bed height state values of MPC 1 and MPC 2

6. Conclusions

A fuel bed height sensor is a critical element in the control of the BioGrate boiler and for optimal energy

production its faulty operation should thus be avoided. In this paper a FTMPC strategy was proposed to

accommodate the fault in the fuel bed height sensor by active controller reconfiguration where two different

control configurations are run in parallel. In these configurations, two alternative control variables, fuel bed

height and combustion power, were utilized.

The FTMPC was tested with the simulated BioPower 5 CHP plant. On the basis of the simulation results,

the proposed FTMPC was able to counter the most typical fault in the BioPower 5 CHP plant caused by

the unknown fuel quality and the status of the furnace (amount of fuel in the furnace). Therefore, the

performance and the profitability of the BioPower 5 CHP plant would be significantly enhanced if such an

FTMPC strategy is implemented.

References

Bauer R., Grölles M., Brunner T., Dourdoumas N. Obernberger I., 2010, Modelling of grate combustion in

a medium scale biomass furnace for control purposes, Biomass. Bioenerg., 34(4), 417-427.

Boriouchkine A., Zakharov A., Jämsä-Jounela S-L., 2012, Dynamic modeling of combustion in a BioGrate

furnace: The effect of operation paramets on biomass firing, Chem. Eng. Sci., 69(1), 669-678.

Grölles M., Reiter S., Brunner T., Dourdoumas N., Obernberger I., 2014, Model based control of a small-

scale biomass boiler, Control. Eng. Pract., 22, 94-102.

Johansson R., Thunman H., Leckner B., 2007, Influence of intraparticle gradients in modeling of fixed bed

combustion, Combust and Flame., 149(1-2), 49-62.

Kortela J., Jämsä-Jounela S-L., 2012, Fuel-quality soft sensor using the dynamic superheater model for

control strategy improvement of the BioPower 5 CHP plant, J. Electr. Power Energy. Syst., 42(1), 38-

48.

Zhang Y., Jiang J., 2008, Bibliographical review on reconfigurable fault-tolerant control systems, Annu.

Rev. Contr., 32(2), 229-252.

Sourander M., Vermasvuori M., Sauter D., Liikala T., 2009, Fault tolerant control for a dearomatisation

process, J. Process. Contr., 19(7), 1091-1102.

Jain T., Yamé J.J., 2013, A Novel Approach to Real-Time Fault Accommodation in NREL's 5-MW Wind

Turbine Systems, IEEE Transactions on Sustainable ener., 4(4), 1082-1090.

Kettunen M., Jämsä-Jounela S-L., 2011, Data-Based, Fault-Tolerant Model Predictive Control of a

Complex Industrial Dearomatization Process, Ind. Eng. Chem. Res., 50(11), 6755-6768.

Maciejowski J.M., 2002, Predictive Control with Constraints. Harlow, UK: Prentice Hal.