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

net/publication/305737398

Fault-tolerant model predictive control (FTMPC) for the BioGrate boiler

Conference Paper · November 2015

DOI: 10.1109/ETFA.2015.7301478

CITATIONS READS

2 54

2 authors:

Sirkka-Liisa Jamsa-Jounela Jukka Kortela

Aalto University Aalto University
185 PUBLICATIONS   1,266 CITATIONS    16 PUBLICATIONS   84 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Artificial Intelligence applied for industrial lime kiln control View project

PAPYRUS View project

All content following this page was uploaded by Sirkka-Liisa Jamsa-Jounela on 01 August 2016.

The user has requested enhancement of the downloaded file.

 This is an electronic reprint of the original article.
This reprint may differ from the original in pagination and typographic detail.

Kortela, Jukka; Jämsä-Jounela, Sirkka-Liisa

Fault-tolerant model predictive control (FTMPC) for the BioGrate boiler

Published in:
20th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA 2015),
Luxembourg, September 8-11, 2015

Published: 01/01/2015

Document Version
Peer reviewed version

Please cite the original version:

Kortela, J., & Jämsä-Jounela, S-L. (2015). Fault-tolerant model predictive control (FTMPC) for the BioGrate
boiler. In 20th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA 2015),
Luxembourg, September 8-11, 2015. (pp. 1-6). Luxembourg: University of Luxembourg, Luxembourg and IEEE
Industrial Electronics Society.

This material is protected by copyright and other intellectual property rights, and duplication or sale of all or
part of any of the repository collections is not permitted, except that material may be duplicated by you for
your research use or educational purposes in electronic or print form. You must obtain permission for any
other use. Electronic or print copies may not be offered, whether for sale or otherwise to anyone who is not
an authorised user.

Powered by TCPDF (www.tcpdf.org)

Fault-tolerant model predictive control (FTMPC) for
the BioGrate boiler

Jukka Kortela Sirkka-Liisa Jämsä-Jounela

Aalto University School of Chemical Technology Aalto University School of Chemical Technology
PL 16100, FI-00076 Aalto PL 16100, FI-00076 Aalto
Email: [email protected] Email: [email protected]

Abstract—The fuel bed height sensor is a critical element in waste. Gölles et al. [6], [7] implemented and experimentally
the control of the BioGrate boiler. A fault appearing in this verified a model based control in a commercially available
sensor greatly affects the control performance in the sense that small-scale biomass boiler using the simplified first-principle
air distribution in the BioGrate boiler deviates from its nominal model. In more details, the mass of water in the water
distribution. To address this problem, a fault tolerant model evaporation zone and the mass of dry fuel in the thermal
predictive control (FTMPC) has been developed to accommodate
the fault in this fuel bed height sensor by the active controller
decomposition zone on the grate are considered as the states of
reconfiguration. In this fault tolerant strategy, water evaporation the simplified model and are estimated by an extended Kalman
in the furnace is estimated by fuel moisture soft-sensor, and filter. Test results showed that the control was always able to
thermal decomposition of dry fuel is estimated by utilizing oxygen provide the required power whereas the conventional control
consumption. This renders the power output of the boiler to be (PID control based on standard control strategies) could not
accurately predicted and controlled. The proposed FTMPC is tolerate a feed water temperature drop of more than 7 ◦ C. In
successfully tested with the BioPower 5 CHP plant data and the addition, the control was able to operate the plant with a lower
results are presented, analyzed, and discussed. excess oxygen content during the load drop and especially
under partial load conditions. The better control of the residual
I. I NTRODUCTION oxygen and the control of the air ratio led to lower emissions
and higher efficiencies. In addition, the model-based control
The utilization of biomass fuel for heat and power pro- was able to handle without difficulties a step-wise change in
duction is growing due to an increasing demand for the the fuel moisture content from 26% to 38% and vice versa.
replacement of fossil energy sources with renewable energy. However, in addition to controlling the power production, the
As a result, the fast and the efficient control of power pro- plant control has to maintain the optimal operating conditions
ducing units becomes increasingly important in combustion in the furnace. According to the boiler design, for the complete
of biomass [1]. However, the main challenges in biomass combustion of biomass, the fuel bed height should be kept at
combustion control are caused by the unpredictable variability the level to achieve the specified ratio between the primary
of the fuel quality, which results in disturbances, faults, and and secondary air, and the amount of fuel in the furnace [8].
failures in the plant behavior and operations. In particular, this
In this paper a FTMPC strategy is proposed to accommo-
is true for the grate firing that is one of the main technologies
date the fault in fuel bed height sensor by active controller
currently used in biomass combustion [2].
reconfiguration. The paper is organized as follows: Section 2
Several different control strategies have been developed presents the BioPower 5 CHP process. The FTMPC strategy
to control the combustion. The combustion power method is presented in Section 3. The test results are given in Section
developed by Kortela and Lautala [3] was employed by 4, followed by the conclusions in Section 5.
many control strategies to compensate variations in the fuel
quality. Based on the combustion power method, in the same II. D ESCRIPTION OF THE B IO P OWER 5 CHP PROCESS
publication Kortela and Lautala [3] suggested a feed-forward The BioPower 5 CHP process consists of two main parts:
control: adjusting the fuel feed flow according to the thermal the furnace and the steam-water circuit. The heat used for
decomposition rate to stabilize the amount of the fuel in the steam generation is obtained by burning solid biomass fuel
furnace. As a result, the effect of the feed disturbance on the – consisting of bark, sawdust, and pellets – which is fed into
generated steam pressure decreased to about one third of the the furnace together with combustion air. The heat of the flue
original value, and the settling time decreased from 45 min gas is transfered by the heat exchangers to the steam-water
to only 13 min. The same method has later been applied to a circulation, where superheated steam is generated [9].
grate boiler [4].
In the BioGrate system, the fuel is fed onto the center
Recently, the model predictive control has proven to be of a grate from below through a stoker screw, as shown in
a successful method for controlling renewable fuel power Fig. 1. The grate consists of alternate rotating and stationary
plants. In particular, the benefits of MPC-based control over concentric rings with the rotating rings alternately rotated
conventional multivariable control have been demonstrated by clockwise and counter-clockwise by hydraulics. This design
Leskens et al. [5] at a grate boiler combusting municipal solid distributes the fuel evenly over the entire grate, with the
burning fuel forming an even layer of the required thickness.
 III. FTMPC FOR THE B IO G RATE BOILER
The overall structure of the FTMPC follows the active
FTC scheme, adjusting the plant control according to the
fault diagnosis results. In more detail, two different MPC
configurations have been developed for the cases of normal and
faulty operations of the fuel bed height sensor. In the faultless
mode, the MPC configuration is as follows: the primary air
flow rate and the stoker speed are the manipulated variables
(u); 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 (y). The fault is
accommodated by employing an alternative estimation of the
fuel bed height, which is based on the thermal decomposition
rate. However, as the alternative estimation is less accurate, the
control reconfiguration is also needed, shifting its focus to the
Fig. 1. 1. Fuel, 2. Primary air, 3. Secondary air, 4. Economizer, 5. Drum, 6. combustion power control while the fuel height is given a low
Evaporator, 7. Superheaters, 8. Superheated steam priority. Additionally, the fuel bed height is kept within the
security limits in both configurations in order to avoid plant
shutdowns.
The moisture content of the wet fuel in the centre of the In more details, the FTC scheme is presented in Fig. 2. The
grate evaporates rapidly due to the heat of the surrounding combustion power and fuel moisture soft-sensors are used to
burning fuel and the thermal radiation coming from the brick compensate the effect of the fuel quality variations. The results
walls. The gasification and visible combustion of the gases and show that these soft-sensors predict the thermal decomposition
solid carbon takes place as the fuel moves to the periphery of rate of dry fuel and the water evaporation with good precision.
the circular grate. At the edge of the grate, ash finally falls Furthermore, the results of the tests show that these methods
into a water-filled ash basin underneath the grate. are able to detect variations in these properties within seconds
[10]. In particular, the fuel moisture estimation is considered by
the MPC as a measured disturbance and is also used to estimate
The primary air for combustion and the recirculation flue
the amount of water in the furnace. Considering the combus-
gas are fed from underneath the grate and they penetrate the
tion power as a model state enables rapid energy production
fuel through the slots in the concentric rings. The secondary
level changes and improves the control performance during the
air is fed directly into the flame above the grate and the
transitions. In addition, the thermal decomposition rate is used
air distribution is controlled by dampers and speed-controlled
in the calculations of the fuel bed height (estimator 2 in Fig. 2),
fans. The gases released from biomass conversion on the
which makes the fault detection and accommodation possible.
grate and a small number of entrained fuel particles continue
According to the fault detection results, the decision on the
to combust in the freeboard, in which the secondary air
control reconfiguration is made, which is then communicated
supply plays a significant role in the mixing, burnout, and the
to the fault accommodation and the FTMPC. Depending on the
formation of emissions. The design of the air supply system,
rp value, the fault accommodation employs either the fuel bed
the ratio between primary and secondary air, plays a key role
height measurement or the thermal decomposition rate and the
in the efficient and complete combustion of biomass [8]. In
primary air flow for the MPC state estimation. Also, FTMPC
modern grate-fired boilers burning biomass, the split ratio of
is switched between the normal and the faulty configurations
primary to secondary air is 40/60, which should be followed
according to the rp signal.
by a control design for the most efficient energy production.
The overall excess air for most biomass fuels is normally set
at 25% or above. A. Controller reconfiguration
1) Detection of faults in the fuel bed height sensor:
The essential components of the water-steam circuit are Two state estimators in Fig. 2 utilize fuel moisture soft-
an economizer, a drum, an evaporator, and superheaters. Feed sensor and combustion power estimations, steam, temperature,
water is pumped from a feed water tank into the boiler. First drum pressure measurements, and alternatively fuel bed height
the water is led into the economizer (4), which is the last heat measurement and calculated fuel bed height to filter the states
exchanger extracting the energy from the flue gas, and thus, of the system, Fig. 3. In order to detect faults in the fuel bed
improving the efficiency of the boiler. From the economizer, height sensor, its filtered calculated value is compared with
the heated feed water is transferred into the drum (5) and along the filtered measurement (pressure drop over the grate). The
downcomers into the bottom of the evaporator (6) through calculated fuel bed height can be expressed from the primary
tubes that surround the boiler. From the evaporator tubes, the air flow rate and the thermal decomposition rate as follows:
heated water and steam return back into the steam drum, where cthd · ṁpa · βthd − ṁgf
they are separated. The steam rises to the top of the steam mds = (1)
drum and flows into the superheaters (7) where it heats up cds
further and superheats. The superheated high-pressure steam where cthd is the thermal decomposition rate coefficient, ṁpa
(8) is then passed into the steam turbine, where electricity is is the primary air flow rate (m3 /s), βthd is the coefficient for
generated. a dependence on the position of the moving grate, ṁgf is the
 Soft-sensors

Combined
combustion power and
fuel moisture soft-sensor

Fault accomodation /
controller reconfiguration State estimations
FDD

Fuel bed height

measurement

Controller + State Water evaporation

reconfiguration estimator 1 Combustion power
- Steam demand
Drum pressure

Thermal decomposition
rate
Primary air

State Water evaporation

Water in
the fuel feed estimator 2 Combustion power

Steam Steam demand

demand Drum pressure
rp x^ d2 d1

Fuel bed height r1 fy y0

Combustion power r2 Primary air u2 +
FTMPC +
Drum pressure r3 Fuel u1 Process

Fig. 2. FTMPC of the BioGrate boiler

u2
Primary
air

u1 y2 y3
Stoker Combustion power Drum
speed Fuel bed Model of thermal Combustion estimation Drum pressure
x1 -
height decomposition power pressure
x3 x5
model of dry fuel model model
y1 - x4 -
Fuel bed height measurement / d2
d1 Fuel bed height estimation Power demand
Moisture
in fuel Water
x
evaporation 2
model Fuel moisture soft-sensor

Fig. 3. The models of the BioGrate boiler

thermal decomposition rate of the fuel, cds is the fuel bed The failure of the fuel bed height measurement is detected if
height cofficient, describing the mass of the fuel proportional the RM SEP exceeds the detection threshold:
to the density of the fuel. If a bias of magnitude by,i occurs at
time instant t in the ith sensor, then the measurement output v
n
for this sensor is given by [11] uP
u |x̂(i)1,1 − x̂(i)1,2 |2
t
i=1
RM SEP = (5)
y(k) = Cx(k) + v(k) + by,i ey,i σ(k − t) (2) n

Furthermore, when a fuel bed height sensor fault occurs, the where n is the number of the samples in the test data set,
residual ν(k) and the two state fuel bed height estimates x̂(k|k) x̂(i)1,1 is the estimated fuel bed height of the first MPC
start to diverge from each other. configuration, and x̂(i)1,2 the estimated fuel bed height of the
second MPC configuration. The limit of detecting the faults is
set above the normal disturbances of the states. Note that the
ν(k) = y(k) − C x̂(k|k − 1) (3) fault isolation is implicitly done in the above fault detection
x̂(k|k) = x̂(k|k − 1) + K(k)ν(k); x̂(0|0) = x̂(0) (4) procedure.
 2) MPC of the BioGrate boiler: The MPC utilizes the the air necessary for a complete combustion of fuel. From the
linear state space system [12]: amount of dry fuel in the thermal decomposition zone, the
input variable ṁpa and the constant parameters (cthd , βthd ,
x(k + 1) = Ax(k) + Bu(k) + Ed(k) cds ), the set point r1 for the mass of dry fuel in the thermal
y(k) = Cx(k) (6) decomposition zone is calculated.
where A is the state matrix, B is the input matrix, E is the cthd · ṁpa · βthd − ṁgf
matrix for the measured disturbances, and C is the output mds = (18)
cds
matrix. According to (6), the k-step ahead prediction is
formulated as: Two different MPC configurations are developed for the pro-
k−1 cess operating in two different modes, i.e. faultless or healthy
mode and faulty mode. In the faultless mode, the primary air
X
y(k) = CAk x(0) + H(k − j)u(j) (7)
j=0
flow rate and the stoker speed are the manipulated variables
(u); the moisture content in the fuel feed and the steam demand
where H(k − j) contains the impulse response coefficients. are the measured disturbances (d); and the fuel bed height and
Therefore, using the Equation (7), the MPC optimization the steam pressure are the controlled variables (y). While for
problem is: the faulty mode, the controlled variables are modified: i.e. the
Np output y is composed of the fuel bed height, the combustion
1X 1 power and the steam pressure. Once the fault is detected and
min φ = ky(k) − r(k)k2Qz + k∆u(k)k2Qu
2 2 isolated using the scheme described in Section III-A1, the
k=1
controller is reconfigured from the healthy mode to the faulty
s.t.x(k + 1) = Ax(k) + Bu(k) + Ed(k), mode.
k = 0, 1, . . . , Np − 1 (8)
y(k) = Cx(k), k = 0, 1, . . . , Np IV. T EST RESULTS OF THE FTMPC STRATEGY
umin ≤ u(k) ≤ umax , k = 0, 1, . . . , Np − 1
∆umin ≤ ∆u(k) ≤ ∆umax , k = 0, 1, . . . , Np − 1 A. Description of the simulation and testing environment
ymin ≤ y(k) ≤ ymax , k = 1, 2, . . . , Np A simulation model of the BioPower 5 CHP plant was built
in the MATLAB environment and additionally, the code for the
where r is the target value and ∆u(k) = u(k) − u(k − 1). FTMPC was developed. Parameters of the models of the water
The process models of the BioGrate boiler have been evaporation, the thermal decomposition of the dry fuel, and the
developed in [13] and [14]. Defining the inputs u, states x, drum were identified by using the data from the BioPower 5
outputs y and the measured disturbances d according to Fig. CHP plant. Moreover, to identify the fuel bed height model,
3, the process models of the BioGrate are as follows: the plant was further modified by installing 8 pressure sensors
for the BioGrate to measure the fuel bed height pressure.
dx1
= cds x1 − cthd βthd u2 + cds,in u1 + w2 (9)
dt
dx2 B. Test results of the FTMPC strategy
= −cwev βwev x2 + cw,in d1 + w1 (10)
dt
dx3
= −x3 + qwf (cthd βthd u2 − cds x1 ) (11) 1.5
dt
height (m)
Fuel bed

1
−0.0244cwev βwev x2 + w3 (12)
0.5
dx4
= −x4 + d2 (13) 0
dt 0 200 400 600 800 1000 1200
Time (second)
dx5 1
= (x3 − x4 ) + w4 (14) 3
flow (kg/s)

dt e
Dry fuel

2
y1 = x 1 + v1 (15)
1
y2 = x 3 + v2 (16) 0
0 200 400 600 800 1000 1200
y3 = x 5 + v3 (17) Time (second)
4
where cds,in is the correction coefficient identified from the
Moisture in
fuel (kg/s)

data, βwev is the coefficient for a dependence on the position 2

from the center to the periphery of the moving grate, cwev and
0
cw,in are the model parameters estimated from the process 0 200 400 600 800 1000 1200
data, qwf is the effective heat value of the fuel (higher heat Time (second)

value) and 0.0244 the heat of vaporization of water.

The set points r2 and r3 for the combustion power and the Fig. 4. Responses of the moisture in fuel, dry fuel flow, and fuel bed height
drum pressure directly result from procedural considerations. to 100% bias fault in the fuel bed height sensor without the FTMPC active.
The set point for the combustion power is calculated according
to the steam demand and the drum pressure is kept constant. To demonstrate the effectiveness of the proposed FTMPC
An important process parameter is λf b describing the ratio strategy, the performance of the FTMPC was evaluated using
of primary air fed to the fuel bed and minimum amount of the BioGrate boiler simulator in a MATLAB environment.
 50.2 50.2
Pressure (bar)

Pressure (bar)
50 50

49.8 49.8
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Time (second) Time (second)
20 20
power (MW)

power (MW)
Combustion

Combustion
15 15
10 10
5 5
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Time (second) Time (second)
3.5 3.5
Primary air

Primary air
flow (m /s)

flow (m /s)
3 3
3

3
2.5 2.5
2 2
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Time (second) Time (second)

Fig. 5. Responses of the pressure, combustion power, and primary air flow Fig. 7. Responses of the pressure, combustion power, and primary air flow
to 100% bias fault in the fuel bed height sensor without the FTMPC active. to 100% bias fault in the fuel bed height sensor with the FTMPC active.

1.5 0.06
height (m)
Fuel bed

1
0.5 0.05
0
0 200 400 600 800 1000 1200
Time (second)
0.04
3
RMSEP index
flow (kg/s)
Dry fuel

2
0.03
1
0
0 200 400 600 800 1000 1200
Time (second) 0.02
4
Moisture in
fuel (kg/s)

2 0.01

0
0 200 400 600 800 1000 1200 0
Time (second) 0 200 400 600 800 1000 1200
Time (second)

Fig. 6. Responses of the moisture in fuel, dry fuel flow, and fuel bed height Fig. 8. Scenario 1: RMSEP index of fuel bed height state of MPC 1 and
to 100% bias fault in the fuel bed height sensor with the FTMPC active. MPC 2.

The input limits were u1,min = 0, u1,max = 4, ∆u1,min = The test scenario had a downward step-shaped fault in
−0.03, and ∆u1,max = 0.03 [kg/s] for the stoker speed; the fuel bed height measurement of 100% of the nominal
u2,min = 0, u2,max = 4, ∆u2,min = −0.03, and ∆u2,max = value and the power demand was changed from 12 MW to 16
0.03 [kg/s] for the primary air. MW after 200 seconds. The fault was introduced into the fuel
bed height measurement after 500 seconds. Then, the power
In the nominal case, the output limits were y1,min = 0.2, demand was changed from 16 MW to 12 MW during the
y1,max = 1 [m] for the fuel bed height; and y2,min = 0, time period of 800 - 1000 seconds. As it can be seen from
y2,max = 55 [bar] for the drum pressure. the Figs. 4-7, the fault resulted in the extremely high values
    of the primary air and the fuel bed height. Fig. 8 shows the
0.1 0 0.1 0 RMSEP index of the different fuel bed height state of MPC
Qz,1 = and Qu,1 =
0 0.1 0 0.1 1 and MPC 2.
In the reconfiguration, the output limits were y1,min = 0.2,
y1,max = 1 [m] for the fuel bed height; y2,min = 0, V. C ONCLUSION
y2,max = 30 [MW] for the combustion power; and
y3,min = 0, y3,max = 55 [bar] for the drum pressure. A fuel bed height sensor is a critical element in the control
of the BioGrate boiler and for optimal energy production its
0.001 0 0
" #
faulty operation should thus be avoided. In this paper a FTMPC
 
0.1 0
Qz,2 = 0 0.001 0 and Qu,2 = strategy was proposed to accommodate the fault in the fuel
0 0.1
0 0 0.1 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.

R EFERENCES
[1] Edlund, K., Bendtsen, J.D., Jørgensen, J.B. (2011). Hierarchical model-
based predictive control of a power plant portfolio. Control Engineering
Practice, 19(10), 1126–1136.
[2] Leão, R.P.S., Barroso, G.C., Sampaio, R.F., Almada, J.B., Lima, C.F.P.,
Rego, M.C.O., Antunes, F.L.M.(2011). The future of low voltage
networks: Moving from passive to active. International Journal of
Electrical Power & Energy Systems, 33(8), 1506–1512.
[3] Kortela, U., Lautala, P. (1982). A new control concept for a coal power
plant. Control Science and Technology for the Progress of Society, 6,
3017–3023.
[4] Kortela, U., Marttinen, A. (1985). Modelling, Identification and Control
of a Grate Boiler. In Proceedings of the 1985 American Control
Conference. Boston, 19-21 June 1985, pp. 544–549.
[5] Leskens, M., van Kessel, L.B.M., Bosgra, O.H. (2005). Model predictive
control as a tool for improving the process operation of MSW combustion
plants. Waste Management, 25(8), 788–798.
[6] Gölles, M., Bauer, R., Brunner, T., Dourdoumas, N., Obernberger, I.
(2011). Model based control of a biomass grate furnace. In Proceedings
of the 9th European conference on industrial furnaces and boilers.
Estoril, 26-29 April 2011, pp. 1–10.
[7] Gölles, M., Reiter, S., Brunner, T., Dourdoumas, N., Obernberger, I.
(2014). Model based control of a small-scale biomass boiler. Control
Engineering Practice, 22, 94–102.
[8] Yin, C., Rosendahl, L.A., and Kær, S.K. (2008). Grate-firing of biomass
for heat and power production. Progress in Energy and Combustion
Science, 34(6), 725–754.
[9] Boriouchkine, A., Zakharov, A., Jämsä-Jounela, S-L. (2012). Dynamic
modeling of combustion in a BioGrate furnace: The effect of operation
parameters on biomass firing. Chemical Engineering Science, 69(1),
669–678.
[10] Kortela, J., Jämsä-Jounela, S.-L. (2013). Fuel moisture soft-sensor and
its validation for the industrial BioPower 5 CHP plant. Applied Energy,
105, 66–74.
[11] Prakash, J., Patwardhan, S.C., Narasimhan, S. (2002). A Supervisory
Approach to Fault-Tolerant Control of Linear Multivariable Systems.
Industrial & Engineering Chemistry Research, 41(9), 2270–2281.
[12] Maciejowski, J.M. (2000). Predictive Control with Constraints. Prentice
Hall, Harlow, pp. 36–150.
[13] Kortela, J., Jämsä-Jounela, S.-L. (2014). Model predictive control
utilizing fuel and moisture soft-sensors for the BioPower 5 combined
heat and power (CHP) plant. Applied Energy, 131, 189–200.
[14] Kortela, J., Jämsä-Jounela, S.-L. (2015). Modeling and model predictive
control of the BioPower combined heat and power (CHP) plant. Inter-
national Journal of Electrical Power & Energy Systems, 65, 453–462.

View publication stats