Sunteți pe pagina 1din 56

Iranian Journal of Science and Technology, Transactions of Electrical

Engineering
Optimal Design of Proportional-Integral Controllers for Grid Connected Solid Oxide
Fuel Cell Power Plant Employing Differential Evolution Algorithm
--Manuscript Draft--

Manuscript Number: ISTE-D-16-00197R1

Full Title: Optimal Design of Proportional-Integral Controllers for Grid Connected Solid Oxide
Fuel Cell Power Plant Employing Differential Evolution Algorithm

Article Type: Research Paper

Funding Information:

Abstract: This paper proposes the application of Differential Evolution (DE) algorithm for the
optimal tuning of Proportional-Integral controller designed to improve the small signal
dynamic response of a grid connected solid oxide fuel cell (SOFC) system. The small
signal model of the study system is derived and considered for the controller design as
the target here is to track small variations in SOFC load current. The proposed
proportional-integral (PI) controllers are incorporated in the feedback loops of hydrogen
and oxygen partial pressures, grid current d-q components and dc voltage with an aim
to improve the small signal dynamic responses. The controller design problem is
formulated as the minimization of an eigenvalue based objective function where the
target is to find out the optimal gains of the PI controllers in such a way that the
discrepancy of the obtained and desired eigenvalues are minimized. Eigenvalue and
time domain simulations are presented for both open-loop and closed loop systems. To
test the efficacy of DE over other optimization tools, the results obtained with DE are
compared with those obtained by particle swarm optimization (PSO) algorithm and
invasive weed optimization (IWO) algorithm. Three different types of load disturbances
are considered for the time domain based results to investigate the performances of
different optimizers under different sorts of load variations. Moreover, non-parametric
statistical analyses, namely, one sample Kolmogorov-Smirnov (KS) test and Paired
sample t-test are used to identify the statistical advantage of DE algorithm over other
two. The presented results suggest the supremacy of DE over PSO and IWO in finding
the optimal solution.

Corresponding Author: Ashik Ahmed


Islamic University of Technology
BANGLADESH

Corresponding Author Secondary


Information:

Corresponding Author's Institution: Islamic University of Technology

Corresponding Author's Secondary


Institution:

First Author: Ashik Ahmed

First Author Secondary Information:

Order of Authors: Ashik Ahmed

Md Shahid Ullah, PhD

Md Ashraful Hoque, PhD

Order of Authors Secondary Information:

Author Comments: Dear Editor,

According to the feedback provided by the reviewers, the manuscript has been
updated.

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
Manuscript Click here to
access/download;Manuscript;SOFC_grid_PI_control_journal
Click here to view linked References
1
2
3
4
5 Optimal Design of Proportional-Integral Controllers for Grid-
6
7
8 Connected Solid Oxide Fuel Cell Power Plant Employing
9
10
11 Differential Evolution Algorithm
12
13
14
15 Ashik Ahmed
16 Assistant Professor
17 EEE Department
18 Islamic University of Technology
19 Bangladesh
20
21
*ashikhmd@yahoo.com(corresponding author)
22 Tel: +8801864092662
23
24
25 Md. ShahidUllah
26 Professor
27 EEE Department
28 Daffodil International University
29 Bangladesh
30
31
32
33 Md. AshrafulHoque
34 Professor
35 EEE Department
36 Islamic University ofTechnology
37 Bangladesh
38
39
40
41
42
43
44
45 Acknowledgement
46
47
48 The authors wish to express their acknowledgement for the support from EEE Department, Islamic University of
49
50 Technology, Bangladesh in completing this work.
51
52
53
54
55
56
57
58
59
60
61
62 1
63
64
65
1
2
3
4 Abstract - This paper proposes the application of Differential Evolution (DE) algorithm for the optimal tuning of
5
6 Proportional-Integral controller designed to improve the small signal dynamic response of a grid connected solid
7
8 oxide fuel cell (SOFC) system. The small signal model of the study system is derived and considered for the
9
10 controller design as the target here is to track small variations in SOFC load current. The proposed proportional-
11
12 integral (PI) controllers are incorporated in the feedback loops of hydrogen and oxygen partial pressures, grid
13
14 current d-q components and dc voltage with an aim to improve the small signal dynamic responses. The controller
15
16 design problem is formulated as the minimization of an eigenvalue based objective function where the target is to
17
18 find out the optimal gains of the PI controllers in such a way that the discrepancy of the obtained and desired
19
20
eigenvalues are minimized. Eigenvalue and time domain simulations are presented for both open-loop and closed
21
22
loop systems. To test the efficacy of DE over other optimization tools, the results obtained with DE are compared
23
24
with those obtained by particle swarm optimization (PSO) algorithm and invasive weed optimization (IWO)
25
26
algorithm. Three different types of load disturbances are considered for the time domain based results to investigate
27
28
29 the performances of different optimizers under different sorts of load variations. Moreover, non-parametric
30
31 statistical analyses, namely, one sample Kolmogorov-Smirnov (KS) test and Paired sample t-test are used to identify
32
33 the statistical advantage of DE algorithm over other two. The presented results suggest the supremacy of DE over
34
35 PSO and IWO in finding the optimal solution.
36
37
38 Keywords – grid connected solid oxide fuel cell, differential evolution algorithm, small signal model, multi-objective
39
40 optimization, synchronously rotating d-q reference frame.
41
42 I. Introduction
43
44
45 With the progress of science and technology, people are inclined towards increased use of sophisticated equipment
46
47 which asks for more electricity. The fossil fuel based energy resources are diminishing at a rapid rate and thus it
48
49 becomes incumbent to search for sustainable energy sources. At the same time installation of new transmission lines
50
51 to increase transfer of power to remote places is not encouraged due to lack of land access. Introduction of the
52
53 concept of distributed generation (DG) has opened a new paradigm having exciting features like minimum
54
55 transmission loss, less dependence on fossil fuel, steadiness in supply and last but not the least clean in
56
57 nature(Pepermans, Driesen et al. 2005). Among the available DG sources, Fuel Cells (FCs) have observed oneof the
58
59 highest level of application in electricity generation. Solid Oxide Fuel Cell (SOFC), one member of the FC family,
60
61
62 2
63
64
65
1
2
3
4 has recently captured the attention as one of the most efficient electricity producing device (Choudhury, Chandra et
5
6 al. 2013). Application of SOFC in both stand alone and grid connected modes are reported in the literature. The
7
8 stand-alone SOFCs are finding their applications in powering cars, small homes and isolated or distant areas
9
10 whereas the grid connected mode is solely utilized for supplying electric power to the grid.
11
12
13 Integration of SOFC to the grid generally requires intermediate power electronic interface as the power output from
14
15 the cell is purely DC. This interfacing can be achieved through a single stage dc-ac inverter followed by step up
16
17 transformer and filter stages (Jain, Jiang et al. 2012). The transformer stage can be bypassed if dc-dc boost converter
18
19 is used prior to the inverter stage (Du, Wang et al. 2013). In both cases, design of proper control mechanism has to
20
21 be ensured for controlled and stable operation of the whole system. PI controllers have been successfully used for
22
23 both dc-dc converter and dc-ac inverter stages in (Gelen and Yalcinoz 2015) without adoption of any specific
24
25 methodology for selecting the controller parameters. Adaptive hysteresis control of power conditioning unit has
26
27 been proposed in (Bhuyan, Padhee et al. 2014)where the main objective was to improve the dynamic voltage profile
28
29
of SOFC voltage output and the power quality of the system. A two-loop control scheme is proposed in
30
31
(Sedghisigarchi and Feliachi 2006) where a multi-objective optimization problem is solved using genetic algorithm
32
33
(GA). The control target was to keep the fuel utilization and system frequency variation within safe margin for small
34
35
36 change in load. However, the temperature dynamics used in this work was a much simplified one. A novel control
37
38 strategy based on the coordination of hydrogen fuel flow rate and utilization factor has been proposed in
39
40 (Nayeripour and Hoseintabar 2013). Although the results showed improved performance than few other previously
41
42 published works, the proposed system used two dc-dc converters along with a storage device which does not make it
43
44 a cost-effective solution. Separate PI controllers are proposed for controlling SOFC active power and inverter ac
45
46 voltage in (Saha, Chowdhury et al. 2007). An approximated relation between the fuel flow rate and the ac voltage
47
48 phase angle is established for controlling the active power output from the SOFC whereas the ac voltage magnitude
49
50 is controlled by the modulation index. Linearized SOFC dynamic model is used and robust control theory is applied
51
52 for improving disturbance rejection and load following capability of the system in (Knyazkin, Söder et al. 2003).
53
54 Instead of using any proper controller, various control strategies have been defined for different sorts of disturbance
55
56 inputs for controlling the SOFC PCU in (Li, Rajakaruna et al. 2007). The problem with this method lies in the fact
57
58 that the control law has to be switched every time a different type of disturbance is experienced by the system.
59
60 Hierarchical load tracking control mechanism is proposed for ensuring the maximum efficiency operation of the grid
61
62 3
63
64
65
1
2
3
4 connected SOFC in (Li, Wu et al. 2015). The overall control scheme is aimed to track the change in active power
5
6 reference while keeping the temperature within safe limit. Fuzzy logic based control effort is utilized along with two
7
8 other control loops in (Chatterjee, Shankar et al. 2014) for tracking of change in active power demand of the grid
9
10 connected SOFC. Several other fuzzy logic based controller design are reported in (Hajizadeh and Aliakbar-Golkar
11
12 2007, Mammar and Chaker 2009, Bhuyan and Mahapatra 2011, Sudheer and Prasanna 2011). However, due to the
13
14 inherent drawbacks of fuzzy logic based controllers, it has not seen widespread applications (Albertos and Sala
15
16 1998). Optimal PI controllers are designed using differential evolution algorithm for active power control of grid
17
18 connected SOFC system in (Taher and Mansouri 2014). Time domain based objective function is defined and the
19
20
optimal control effort is applied at the dc-ac inverter interface. However, the cell temperature dynamics has been
21
22
completely ignored in the dynamic model of the SOFC.
23
24
25 From the previous works presented above, it becomes evident that very little works have been carried out based on
26
27 the small signal model of the grid connected SOFC system. Analysis of small signal model gives an insight to the
28
29
dynamic behavior of the system from a small disturbance point of view. Participation factor helps in identifying the
30
31
basic relationship between the states and the corresponding eigenvalues and thus helps in deciding the critical
32
33
system variables which must be controlled for system stability (Kundur, Balu et al. 1994). The small signal model
34
35
36 also allows the application of simple linear control techniques for the improvement of dynamic response of the
37
38 system. However, selection of proper control parameters for the optimal response still remains as a challenge.
39
40 In this paper, the temperature dynamics (Huang, Qi et al. 2013) and fuel cell electrical dynamics (Du, Wang et al.
41
42 2012) have been included in the SOFC dynamic model. The inclusion of the temperature dynamics modifies the
43
44 species dynamics and the SOFC output voltage expressions a little. The SOFC is interfaced to the grid with the help
45
46 of a dc-dc converter and a dc-ac inverter stage. The inverter is modeled in the d-q coordinate of a synchronously
47
48 rotating reference frame. As the target in this work is to design controllers which should work well under small
49
50 disturbance, the overall system model is linearized using Taylor series expansion technique (Nise 2011) and the
51
52 disturbance is simulated as small variation in the load. The controllable parameters of the grid connected SOFC
53
54 under consideration are i) the partial pressure of hydrogen gas, ii) partial pressure of oxygen gas, iii) the dc-dc
55
56 converter output voltage which in turn controls the SOFC output voltage, iv) the inverter output active power and v)
57
58 the inverter output reactive power. Conventional PI controllers are fed with the errors between the respective actual
59
60 and reference values and the controller gains are optimized based on aneigenvalue based performance index. The
61
62 4
63
64
65
1
2
3
4 optimal tuning of the controller parameters are performed with the help of differential evolution (DE) algorithm
5
6 which has already found its application in different engineering optimization problem and regarded as one of the
7
8 best (Vesterstrøm and Thomsen 2004, El Ela, Abido et al. 2009, Qin, Huang et al. 2009, Wang, Chiou et al. 2009).
9
10 The DE based results are then compared with those obtained by PSO and IWO algorithms. Two non-parametric
11
12 statistical tests namely, one sample Kolmogorov-Smirnov (KS) test and Paired sample t-test are also used to support
13
14 the findings of the work.
15
16 The mathematical model of the grid connected SOFC is presented in section II. Section III discusses the
17
18 linearization of the model based on the Taylor series expansion and presents the open loop responses. A detail
19
20
description of the DE algorithm is presented in section IV and formulation of the optimization problem is also
21
22
shown. The simulation results showing the effectiveness of the proposed method of controller design are displayed
23
24
in section V and lastly the paper is concluded in section VI.
25
26
27
28
29 II. Mathematical model of grid connected SOFC
30
31 The schematic diagram of the overall system is presented in figure 1. Hydrogen and oxygen are fed into the SOFC to
32
33 generate electricity in the form of output voltage Vfc. The flow rates of hydrogen and oxygen are represented by nh2in
34
35 and no2in, respectively. Depending on the connected load, the amount of cell output current Ifc will vary. The dc-dc
36
37 converter stage boosts up the SOFC output voltage to Vdc which can be manipulated by the duty ratio dc. A dc link
38
39 capacitor Cdc is used for stabilizing the output voltage of the converter stage. Under steady state, the current through
40
41 Cdcis zero and thus the output current of the dc-dc converter stage Idc2 and the input current of the inverter stage Idc1
42
43 becomes equal. The switching pulses are used to trigger the gates of the switching devices in the inverter. Variation
44
45 in the switching pulses will introduce controlling effect on the inverter output Vabc. The grid filter comprising of Rf
46
47 and Lf takes care of any unwanted harmonics in Vabc. The grid is represented by three phase balanced voltage source
48
49 eabc and the inverter output current is denoted as iabc. The defining mathematical expressions for each block are given
50
51 in the following subsections.
52
53
54
55 Fig. 1 Grid connected SOFC system
56
57 SOFC:
58
59 The reduction-oxidation (Red-Ox) reaction in the SOFC is presented in equation (1–3).
60
61
62 5
63
64
65
1
2
3
4 1
5 Cathode reaction (reduction of oxygen): O2  2e   O 2 (1)
6 2
7
8
2
9 Anode reaction (oxidation of hydrogen): H 2  O  H 2O  2e  (2)
10
11
12
1
13 Overall reaction: H 2  O2  H 2O (3)
14 2
15
16
17 The cell output voltage is expressed by Nernst equation which is given by:
18
RT  PH 2 PO2 
19 0.5
20 E  E0  ln   (4)
21 2 F  PH 2O 
22
23
24
25 Here, PH 2 , PO2 and PH 2O are the partial pressures of the hydrogen, oxygen and water vapor, respectively. R is the
26
27 universal gas constant in atm/(mol-K), T is the electrode temperature in kelvin, F is the Faraday’s constant in C/mol
28
29 and E0 is the SOFC open circuit voltage which is a function of gibbs’ free energy of eq. (3). For an SOFC stack with
30
31 N0 number of cells in series, the SOFC voltage becomes:
32
33
34
N 0 RT  PH 2 PO2 
0.5
35
V  N 0 E0  ln 
 PH O 
36 (5)
2F  
37 2

38
39
40 Here, N 0 represents the number of cells in the stack connected in series. The open circuit voltage (OCV) E 0 is not a
41
42
43 constant but a function of cell temperature as reported in (Campanari and Iora 2004).This temperature dependence
44
45 can be expressed as: E0  Enoloss  0.000252T (6)
46
47
48 where Enoloss is the theoretical maximum voltage of the cell. Under loaded condition, the ohmic drop can be
49
50
51 represented as a function of cell operating temperature as (Sedghisigarchi and Feliachi 2004):
52
53
54   1 1 
55 Eohmic  r0 exp      I (7)
56   T T0  
57
58
59
60
61
62 6
63
64
65
1
2
3
4
5
Here, r0 : resistance at standard temperature T0 ,  : constant coefficient, I : cell current at the operating
6
7 temperature T. Inclusion of this loss modifies the expression of cell voltage as:
8
9
N 0 RT  PH 2 PO2    1 1 
10 0.5
11 Vs  N 0 E0  ln    r0 exp      I (8)
 PH O 
12 2F  2    T T0  
13
14
15
16 The SOFC power can simply be expressed as: Psofc  Vs I (9)
17
18
19 Equation (8-9) governs the steady state characteristics of a SOFC.
20
21
22 SOFC dynamic model:
23
24
25 Two major balances are to be considered to represent the overall dynamics of an SOFC. These are – component
26
27 material balance and energy balance. The component material balance represents the change in balance which
28
29 occurs in different species during the chemical reaction in an SOFC whereas the energy balance ensures the
30
31 equilibrium between associated input and output energy levels.
32
33
34 Component material balance:
35
36
37 Considering inflow, outflow and reaction rates of the SOFC, the partial pressure dynamics for the i-th species of
38
39 SOFC can be expressed as:
40
41
42 V
43
Pi  niin  nir  niout (10)
RT
44
45
46 in out r
47 where ni , ni and ni are the inlet, outlet and reactive molar flow rates for i-th species and V is the volume of
48
49 anode or cathode chamber. If the pressure drop in the channel between the inlet and outlet is assumed negligible
50
51 with a choked orifice (Murshed, Huang et al. 2007), the outlet and reactive flow rates can be written as
52
53
54 nir  nm K r I (11)
55
56
57
58 niout  Ki Pi (12)
59
60
61
62 7
63
64
65
1
2
3
4
where nm is the number of moles of a certain specie present in the reaction, K r = N0/(4F), Ki and Pi are the valve
5
6
7 molar constant and partial pressure of i-th specie. The nm value for hydrogen, oxygen and water (vapor) are 2, 1 and
8
9 V
10 2 respectively as per equation (3). Using equations (11) and (12) and defining i  ,equation (10) can be
11 K i RT
12
13 represented as:
14
15
16
17 Pi 
1
K i i
 niin  nm K r I  Ki Pi  (13)
18
19
20
21  i0T 0
22 Noting that K i is constant and redefining  i  , equation (13) can be expressed as:
23 T
24
25
26 Pi 
T
 T Ki
0
0  niin  nm K r I  K i Pi  (14)
27 i

28
29
30 Applying this generalized relation of equation (14) to the components of SOFC, the following expressions can be
31
32 obtained:
33
34
35
36
PH 2 
T
 T K H2
00 
nHin2  2 K r I  K H 2 PH 2  (15)
H2
37
38
39
40 PO2 
T

nin  K r I  KO2 PO2
 T 0 KO2 O2
0  (16)
41 O2

42
43
T
 
44
45 PH 2O  nHin2O  2 K r I  K H 2O PH 2O (17)
46  0 0
T K H 2O
H 2O
47
48
49 Energy balance:
50
51
52 The energy balance deals with various heat transfers occurring at different layers of an SOFC. If lumped model of
53
54 SOFC is considered, application of first law of thermodynamics around the entire SOFC yields the following
55
56 dynamics of electrode temperature, T
57
58
59
60
61
62 8
63
64
65
1
2
3
4 in T T
dT
5 meC p   niin  C p ,i T  dT   niout  C p ,i T  dT  nHr 2 Hˆ ro  Vs I (18)
6 dt Tref Tref
7
8
9
10 where me and C p are the mass and average specific heat of the cell excluding gases; C p ,i is the specific heat of
11
12
13 ˆ o is the specific heat of reaction and
the i-th fuel or gas entering or leaving the cell; H Vs is the cell stack voltage.
r
14
15 The expressions of specific heats are usually adopted from standard thermodynamics table (Felder and Rousseau
16
17
18 2008) and in general can be expressed as: C p,i (T )  ai  bT
i  ciT  diT
2 3

19
20
21 SOFC electrical dynamics:
22
23
24 The electrical dynamics represent the chemical reaction inside the SOFC to restore the charge which has been used
25
26 up by the load. A first-order transfer function is used to model the dynamic with the time constant Tel . The
27
28 differential form of this dynamics can be expressed as:
29
1
 I fcref  I fc 
30
31 I fc  (19)
32 Tel
33
34 Here, I fcref is the reference current of the SOFC. The ODEs of equation (15-19) and the algebraic equations (8-9)
35
36
37 form the complete differential-algebraic model of the SOFC under consideration.
38
39 DC-DC converter:
40
41 The dc-dc boost converter is modeled by the following algebraic equations:
42
43 V fc
44 Vdc  (20)
45
1  dc
46
47 I dc 2  1  dc  I fc (21)
48
49
50 DC capacitor dynamics:
51
52 Considering the current through the dc capacitor Cdc to be zero under steady state, the dynamics of the dc voltage
53
1
54
55 can be written as: Vdc   I dc 2  I dc1  (22)
Cdc
56
57
58 The inverter input current can be related to the output current in terms of the three phase switching signals Kabc as:
59
60 I dc1  kaia  kbib  kcic (23)
61
62 9
63
64
65
1
2
3
4 Considering balanced system, eq. (23) can be rewritten in d-q coordinate as:
5
6
7
I dc1  kd id  kq iq (24)
8
9 where kdq  Tabc
dq
kabc and idq  Tabc
dq dq
iabc andabc-to-dq transformation matrix Tabc is given as:
10
11
 
   
12
2 cos t  2
13 cos t cos t  3 3 
2 
14
15
16
dq
Tabc   sin t sin t  2
3
3  sin t  2  
3 
(25)
17  1 1 1 
18  2 2 2 
19
20
21 Replacing the expressions of I dc1 and I dc 2 in eq. (24), the following is obtained:
22
23
1  d c  I fc   kd id  kq iq  
1
24 Vdc  (26)
25 Cdc  
26
27
DC-AC inverter:
28
29
Considering PWM based full bridge inverter operation, the relation between inverter input and output voltages can
30
31
be expressed with the help of switching matrix as:
32
33
34 Va   2 1 1  ka 
35 V   Vdc  1 2 1  k  (27)
36  b 3   b
37 Vc   1 1 2   kc 
38
39
Grid voltage eabc can be represented in terms of inverter output voltageVabc as:
40
41
d
42 eabc  Vabc  iabc R f  L f iabc (28)
43 dt
44
45
46
Applying abc-to-dq transformation to eq. (28) and using the expression of Vabc from eq. (27), the following can be
47
48 obtained:
49
50 did R 1 V
51   f id  ed  iq  dc kd (29)
52 dt Lf Lf Lf
53
54 diq Rf 1 V
55  iq  eq  id  dc kq (30)
56 dt Lf Lf Lf
57
58 where  is the angular speed of the synchronously rotating reference frame.
59
60
61
62 10
63
64
65
1
2
3
4 Grid active and reactive power:
5
6 The active and reactive powers of the grid can be represented in d-q coordinate as:
7
8
9
Pgrid  eq iq  ed id (31)
10
11 Qgrid  eq id  ed iq (32)
12
13
14 For decoupling control of Pgridand Qgrid,it is considered that the d-component of grid voltage is aligned with the grid
15
16 voltage phasor e as shown in figure 2. If it is desired to supply power at unity power factor to the grid, from eq. (32)
17
18 it is evident that the q-component of grid current should be set to zero. Under this condition, the real and reactive
19
20 power expressions are modified as:
21
22 Pgrid  ed id (33)
23
24
25 Qgrid  ed iq (34)
26
27
28
29
30 Fig. 2 Grid voltage phasor aligned along the d-axis
31
32
33
34 So, from eq. (33-34) it can be concluded that, decoupled control of active and reactive power can be implemented
35
36 through the control of d and q components of grid current, respectively if the grid voltage can be kept along the d-
37
38 axis. In practice, phase locked loop (PLL) is used for perfect tracking of the grid voltage angle. However, in this
39
40 work the PLL dynamics has not been included as grid disturbance or faults are excluded from the study.
41
42 III. Linearization and open loop response of the study system
43
44
45 The dynamic model of the grid connected SOFC system developed in section II can be represented in state space
46
47
48 form as x  f ( x )  g ( x ).u , where f ( x ) and g ( x ) are nonlinear functions of the state vector x , uis the
49
50
51 control vector, x  [ Ph 2 Po 2 Ph 2o T I fc id iq Vdc ] and u   nh 2in no 2in kd kq d c  . The linearized form of the
52
53
54 state space model will be in the form x  Ax  Bu , where the matrices A and B are listed in the appendix.
55
56 After application of Taylor series expansion and approximation, the state space model in linearized form become:
57
58
59
60
61
62 11
63
64
65
1
2
3

 
4
5 T 0 K h2 Ph02  nhin2 0  2 I 0fc K r 2 K rT 0 T0
Ph2  P  T  I  nhin2 (35)
T0 H0 2 K h2 T0 H0 2 K h2 T0 H0 2 T0 K h2 H0 2
6 h2 fc
7
8
9
10
11
PO2 
T 0
P 

KO2 PO02  nOin20  I 0fc K r
T 
K rT 0
I
 
T0
nOin2 (36)
T0 O2 KO2 T0 O2 KO2 T0 O2 T0 KO2 O2
12 0 O2 0 0 fc 0
13
14
15
16
17 Ph2O 
T 0
P 

K h2O Ph02O  2 I 0fc K r
T 

2 K rT 0
I fc (37)
T0 h02O K h2OT0 h02O K h2OT0 h02O
h2O
18
19
20
21
22 T  Tph Ph2  Tpo PO2  TpwPh2O  Tpt T  Tpi I fc  Tpnh nhin2  Tpno nOin2 (38)
23
24
25
26
27
I fc 
1
Tel
 I fcref  I fc  (39)
28
29
30
Rf kd 0 V
31 id   id  0 iq  Vdc  dc 0 kd (40)
32 Lf Lf Lf
33
34
35
Rf kq 0 Vdc 0
36 iq   iq  0 id  Vdc  k q (41)
37 Lf Lf Lf
38
39
40
41 Vdc  
 dc 0  1 I 
kd 0 k
id  q 0 iq 
I fc 0 i i
d c  d 0 kd  q 0 kq (42)
42
43
C dc I fcb 
fc
Cdc Cdc  Cdc I fcb  Cdc Cdc
44
45
46 The linearizing constants of eq. (38) are listed in the appendix. Ifcb in eq.(42) is the base value of the SOFC output
47
48 current.
49
50
Reference generation:
51
52
53 The change in the reference value of the SOFC current should yield a change in the reference values of the
54
55
controllable variables. If the left hand sides of eq. (17-18) are equated to zero, the resulting steady state relationships
56
57
in linearized form are obtained as:
58
59
60
61
62 12
63
64
65
1
2
3
4
2 K r
5 Ph 2 r   I fcref (43)
6 Kh2
7
8
9
Kr
10 PO 2 r   I fcref (44)
11 KO 2
12
13
14 For generating the d-axis grid current reference, it is identified that for lossless power electronics interface, the
15
16
17 SOFC output dc power and grid side ac power are same, i.e. Psofc  Pgrid . From eq. (9) and (33) it can be shown that
18
19
Vs I fcref
20 idr  . Linearizing this expression the following relation is established:
21 e V
d fcb I fcb 
22
23
24
Vs
25 idr  I
 edV fcb I fcb  fcref
(45)
26
27
28
29
Again, using the relation between Psofc and Pgridalong with eq. (20), the relation between dc bus voltage reference
30
31
32 Vdcr and SOFC current reference I fcref can be established as follows:
33
34
35
36   Pfcpu I fcb 
Vdcr    I fcref
I fcref  1  d c 0  I fcref
37 (46)
38 
39
40
41 The reference value for the q-axis grid current is obtained using eq. (34) and written as:
42
43
44 Qgrid
45 iqr   (47)
46 ed
47
48
49 Open loop response:
50
51
52 The open loop response of the linearized dynamic model is studied and the results are presented in figure 3a-3b.
53
54 Numerical data used for the simulation is listed in the appendix. The dynamics is initiated by changing the fuel cell
55
56 reference current and grid reactive power reference by +30 A and +0.2 pu, respectively. From figure 3a it is found
57
58 that although the hydrogen and oxygen partial pressures track the new reference, it takes a long time to settle down.
59
60 The water partial pressure is found to be rising whereas the temperature dynamics almost settles after 100 sec.
61
62 13
63
64
65
1
2
3
4 Fig. 3a Open loop response of changes in SOFC partial pressures and temperature for a step change in reference
5
6
7 Fig. 3b Open loop response of changes in SOFC current, grid current and dc voltage for step change in reference
8
9 values
10
11
12 From figure 3b it is observed that the SOFC current tracks the reference value within 30 sec. However, the grid
13
14 current d-q components and the dc link voltage are unable to track their respective reference values and moves away
15
16 from the reference with increasing time. Thus, it can be said that proper control action is required for improving the
17
18 dynamic response of the study system. The open loop system eigenvalues are listed in Table 1.
19
20
21 Table 1 Open loop eigenvalues
22
23
24 The participation factor (Kundur, Balu et al. 1994) column of Table 1 reveals that other than the grid current d-q
25
26 component, most of the states are decoupled. As the eigenvalues linked to idq are on the imaginary axis, the system is
27
28 overall marginally stable. Moreover, improvement in the dynamic responses of hydrogen and oxygen partial
29
30 pressures and dc capacitor voltage are also needed for ensuring safe margin of stable operation for the system. To
31
32 improve the small signal dynamic stability of the study system, application of PI controllers are proposed in this
33
34 work whose configurations are shown in figure 4. The controller gains are tuned for an optimal result using DE
35
36 algorithm which is discussed next.
37
38
39 Fig. 4Controller configuration for the grid connected SOFC system of fig. 1
40
41
42
43
44
45 IV. Differential Evolution and the closed loop system
46
47
48 DE is one of the members of evolutionary algorithm family having attractive features of solving multi-objective
49
50 constrained optimization problem. The main steps in DE are- initialization of a set of solution, mutation,
51
52 recombination and selection. In the initialization phase, a random set of probable solution for each parameter is
53
54 generated within the search space. If an objection function with D real parameters is to be optimized for an initial
55
56 population of N, the parameters vector takes the form xi ,G   x1,i ,G , x2,i ,G , ... xD ,i ,G  with i  1, 2,... N and G
57
58
59 is the generation number. With the upper and lower bounds for each parameter defined as x Lj  x j ,i ,1  xUj , the
60
61
62 14
63
64
65
1
2
3
4
random parameters in each generation should satisfy the interval  x j , x j  . Three vectors xr1,G , xr 2,G , and xr 3,G
L U
5
6
7 are randomly selected from a given parameter vector xi ,G for the mutation phase and while doing so, it must be
8
9
10 ensured that the indices r1, r2 and r3 are distinct. These three vectors along with mutation factor M F are used to
11

 
12
13 generate the donor vector, vi ,G 1  xr1,G  M F xr 2,G  xr 3,G
14
15
16
A more useful way of generating vi ,G 1 in the mutation phase is to introduce another variable λ, which changes the
17
18
19 expression of the donor vector as:
20
21
22 vi ,G 1  xi ,G  M F  xr1,G  xr 3,G     xbest ,G  xi ,G  (48)
23
24
25
26 In the recombination phase, trial vector u j ,i ,G 1 is generated which gets updated by the donor vector having
27
28
29 probability CR .
30
31
32  v j ,i ,G 1 if rand j ,i  CR or j  I rand
33 u j ,i ,G 1   (49)
34  x j ,i ,G if rand j ,i  CR and j  I rand
35
36
37 Here, rand j ,i is a random number within the range and I rand is a random integer chosen from [0 1]. In the
38
39
40 selection phase, a comparison is made between the target vector and the trial vector and the one with the best value
41
42 is selected and forwarded to the next generation.
43
44
45
46
ui ,G 1 if J (ui ,G 1 )  J ( xi ,G ) (50)
47 xi ,G 1  
48 xi ,G otherwise
49

50
51
The mutation, recombination and selection phases continue until a pre-specified stopping criterion is fulfilled.
52
53
54
Objective function:
55
56
57
58
59
60
61
62 15
63
64
65
1
2
3
4 The enhancement of small signal stability for the grid connected SOFC is formulated as a multi-objective
5
6 optimization problem. The DE is employed to minimize a multi-objective fitness function comprising of eigenvalue-
7
8 based functions for the PI controller design problem. These objectives are defined as:
9
10
11 n
J1    i   0 
2
12 (51)
13 i 1
14
15
16 n
J 2    0   i 
2
17 (52)
18 i 1
19
20
21 J  ( J1  w * J 2 ) (53)
22
23
24
25 Here,  i is the real part of the ith eigenvalues,  0 is the desired value of the real part of the eigenvalues,  i and  0
26
27 are the actual and desired values of the damping ratios, respectively and w is the weighting factor. Optimization of
28
29 J1 will ensure that the real parts of the eigenvalues are lying near the desired location and that of J 2 will make
30
31
32 sure that the sufficient damping has been injected to the system dynamics. Finally, minimization of J should give
33
34 the optimized value of the controller gains.
35
36
37 The closed loop system:
38
39
40 Once the controllers are incorporated into the system, it turns out into a closed loop system. The additional state
41
42 equations generated due to the controllers are given below:
43
44
45 nhin2  H y1Ph2  H y 2T  H y 3I fc  H y 4nhin2  H y 5Ph 2r (54)
46
47
48 nOin2  Ox1PO2  Ox 2 T  Ox3I fc  Ox 4 nOin2  Ox5PO 2r (55)
49
50
51 Kd  I d1id  I d 2iq  I d 3Vdc  I d 4Kd  I d 5idr (56)
52
53
54
Kq  I q1id  I q 2iq  I q3Vdc  I q 4Kq  I q5iqr (57)
55
56 dc  V1I fc  V2 id  V3iq  V4 Vdc  V5K d  V6d c  V7 Vdcr (58)
57
58
59
60
61
62 16
63
64
65
1
2
3
4
5
Detail expressions of the linearizing constants H y1  H y 5 , Ox1  Ox 5 , I d 1  I d 5 , I q1  I q 5 and V1  V7 are listed
6
7 in the appendix. Inclusion of eq. (54-58) increases the order of the system matrix which now becomes dependent on
8
9 the PI controller gains. The closed loop system in linearized form can be represented as:
10
11 xcl  Asys xcl  Bsysudist (59)
12
13
14
where, x cl   Ph 2 PO 2 Ph2O T I fc id iq Vdc K d K q d c 
T
15 is the closed loop state vector
16
17
T
and udist   I fcref nh2 r nO2 r idr iqr Vdcr 
18 in in
is the disturbance input vector. The matrices Asys and Bsys
19
20
21 are listed in the appendix and it is found from Asys that the system eigenvalues now become function of controller
22
23
gains. So, there is every possibility that for a certain set of controller gains, the closed loop eigenvalues as well as
24
25
the system small signal dynamic response will give optimal response compared to the open loop case.
26
27
28 V. Simulation results
29
30
31 To show the effectiveness of the proposed method of designing controllers for the grid connected SOFC system, the
32
33 performance of the DE algorithm is compared with two other optimization techniques namely, PSO and IWO. Detail
34
35 of the SOFC parameters used in the simulation is presented in the appendix. Eigenvalue analysis and time domain
36
37 based simulations along with two statistical tests are conducted to verify the effectiveness of DE. Details of different
38
39
disturbance scenarios considered for the time domain based simulation are listed in Table 2.
40
41
42
Table 2Detail of different disturbances
43
44
45
46 Controller tuning results:
47
48
49 The result of the proposed tuning method is compared with those obtained by PSO and IWO. The above mentioned
50
51 PI controller tuning procedure was carried out by a computer program coded in MATLAB. The programs were
52
53 executed on a 2.50 GHz Intel Core i5 processor with 4 GB of random access memory (RAM). The parameters used
54
55 for different algorithms are listed in Table 3. From practical point of view, the controller gains should not be too
56
57 high or too low and due to this fact the optimizer search space have been kept within some upper and lower limits
58
59
60
61
62 17
63
64
65
1
2
3
4 which are presented in Table 4. This limited search space should certainly reduce the computational time taken by
5
6 the optimizers.
7
8
9 Table 3 Parameters used for PSO, DE and IWO algorithm
10
11 Table 4 Upper and lower bounds of the controller gains
12
13
14
15
16 Eigenvalue analysis:
17
18
19 For statistical comparison, the developed code was run for 30 times for all the algorithms and the eigenvalues
20
21 obtained for the best outcome are listed in Table 5. Eigenvalues related to the target variables are represented in bold
22
23 format. From Table 5it is seen that under closed loop system,different number of complex pair of eigenvalues are
24
25 obtained and most of their real parts are more negative than the corresponding open loop eigenvalues which ensures
26
27
overall improvement of system dynamic response. Specifically, the real parts of the open loop eigenvalues
28
29
associated with hydrogen and oxygen partial pressures are -0.0383 and -0.3436; whereas the real part of
30
31
corresponding closed loop eigenvalues with DE are -0.2640 and -0.2654, with PSO are -0.2065 and -0.1963, and
32
33
with IWO are -0.2364 and -0.4159, respectively.
34
35
36
37
Table 5 Eigenvalues obtained by PSO, DE and IWO (best from 30 runs)
38
39 Again, in open loop there were no real parts for the d-q components of grid current and thus staying on the
40
41 imaginary axis, whereas under closed loop, the real parts with DE are -1.9344 and -0.1042, with PSO are -1.5683
42
43 and -0.1684 and with IWO are -1.4547 and -0.1670, respectively. In case of dc voltage dynamics DE shows best
44
45 response with an eigenvalue -0.6889 while in open loop it was almost zero. However, as stated earlier, the other
46
47 three eigenvalues associated with temperature, water vapor partial pressure and SOFC current are unchanged even
48
49 with the inclusion of the controllers. The optimized parameters for the best run of all the optimizers are listed in
50
51 Table 6. Observation of Table 6 shows that only DE can obtain all the optimized parameters without hitting any of
52
53 the limits (upper or lower), but PSO reaches the limit in four instances whereas IWO reaches the limit in one
54
55 instance. This shows the ability of DE algorithm to successfully search for an optimal solution within a predefined
56
57 search space.
58
59
60 Table 6 Optimized parameters by PSO, DE and IWO algorithms (best from 30 runs)
61
62 18
63
64
65
1
2
3
4 Time domain based simulation:
5
6
7 The performance of three optimization algorithms for different types of load variations are compared and the
8
9 corresponding time domain simulations are presented in figures 5-19.
10
11
12 Step change in Ifcref:
13
14
15 Figure 5 presents the hydrogen partial pressure responses for step load change. As the eigenvalues suggest, there is
16
17 not much difference identified in the responses of PSO, DE and IWO. None of the responses show any undershoot
18
19 and all settle to the new reference value within 30 sec. Figure 6 shows the oxygen partial pressure responses for step
20
21 load change. In this case it is found that the IWO response gives maximum undershoot while all of the responses
22
23 settle within 25 sec.
24
25
26 Fig. 5 Closed loop response of changes in hydrogen partial pressures for step change in reference values
27
28
29 Fig. 6 Closed loop response of changes in oxygen partial pressures for step change in reference values
30
31
32
33
34
35 Figure 7 shows the d-axis grid current variation for step load change. The PSO and IWO responses show little
36
37 overshoot while the DE responses shows none. Settling time for IWO is found to be the quickest of the three. Figure
38
39 8 shows the q-axis grid current variation for step load change. Study of this figure reveals that both PSO and IWO
40
41 responses show substantial undershoot while the DE shows none. Identical settling timesare observed for the three
42
43 responses.The dc voltage dynamics for step change in references is depicted in figure 9. In this case it is seen that
44
45 the responses of PSO and DE settle much quicker than the IWO response. Specifically, the settling time for PSO and
46
47 DE is around 11.52 sec while for IWO it is 26.27 sec.
48
49
50 Fig. 7 Closed loop response of changes in d-axis grid currentfor step change in reference values
51
52
53 Fig. 8 Closed loop response of changes in q-axis grid currentfor step change in reference values
54
55
56 Fig. 9 Closed loop response of changes in dc voltagefor step change in reference values
57
58
59
60
61
62 19
63
64
65
1
2
3
4 Pulsed change in Ifcref:
5
6
7 A pulse type change in SOFC reference current is applied as mentioned in Table 2. The hydrogen partial pressure
8
9 responses with three different algorithms are shown in figure 10. It is evident from this figure that the PSO response
10
11 gives a little larger undershoot compared to those of DE and IWO while the settling time is almost identical. Figure
12
13 11 shows the variation in oxygen partial pressure for different algorithms. In this case it is found that IWO and PSO
14
15 responses are identical while the DE response has little overshoot before settling to the original reference.
16
17
18 Fig. 10 Closed loop response of changes in hydrogen partial pressure for pulsed change in reference
19
20
21 Fig. 11 Closed loop response of changes in oxygen partial pressure for pulsed change in reference values
22
23
24 Response of d-axis grid current is presented in figure 12 where it is observed that once again the IWO and PSO
25
26 show identical responses while the DE takes littler longer to return to the original reference. The q-axis grid current
27
28 variation for pulsed type reference change is shown in figure 13. In this case, it is seen that the DE based response
29
30 does not show any undershoot whereas substantial undershoot as well as little overshoot is observed for IWO and
31
32 PSO responses. The settling time is, however, almost same for the three. The dc voltage dynamics presented in
33
34 figure 14 shows similar sort of variation for three algorithms.
35
36
37 Fig. 12 Closed loop response of changes in d-axis grid currentfor pulsed change in reference values
38
39
40 Fig. 13 Closed loop response of changes in q-axis grid currentfor pulsed change in reference values
41
42
43 Fig. 14 Closed loop response of changes in dc voltagefor pulsed change in reference values
44
45
46 Staircase type change in Ifcref :
47
48
49 The hydrogen partial pressure response for staircase type change in SOFC reference current is presented in figure
50
51 15. In this case, IWO and DE show identical variations whereas the PSO gives slightly different response. Settling
52
53 times for the three cases are found identical. Slight overshoots and undershoots are found in the oxygen partial
54
55 pressure response for all the algorithms as depicted in figure 16. It is also identified that IWO and PSO gives the
56
57 largest and smallest amount of overshoot andundershoot, respectively.
58
59
60 Fig. 15 Closed loop response of changes in hydrogen partial pressures for staircase type change in reference values
61
62 20
63
64
65
1
2
3
4 Fig. 16 Closed loop response of changes in oxygen partial pressures for staircase type change in reference values
5
6
7 The variations in d-q components of grid current for staircase type change in references are shown in figures 17-18.
8
9 In case of d-axis current response,the performances are found comparable whereas for the q-axis current the DE
10
11 outperforms the other two in terms of undershoot and overshoot. Settling times are found to be identical. With
12
13 regard to the dc voltage dynamics, figure 19 shows that IWO response takes longer to settle to the new references
14
15 while the other two takes almost equal amount of time.
16
17
18 Fig. 17 Closed loop response of changes in d-axis grid currentfor staircase type change in reference values
19
20
21 Fig. 18 Closed loop response of changes in d-axis grid currentfor staircase type change in reference values
22
23
24 Fig. 19 Closed loop response of changes in dc voltagefor staircase type change in reference values
25
26
27 The overall analysis of the time domain based simulations does not give any clear indication about the supremacy of
28
29 one optimization algorithm over the other for the problem under consideration. Hence, to identify the best among the
30
31 three, non-parametric statistical analysis tool isutilizedand the outcome is discussed next.
32
33
34 Comparative study and Non parametric statistical analysis:
35
36
37 A comparative study of DE, PSO and IWO algorithm is presented in Table 7 in terms of elapsed time and best
38
39 fitness value achieved in minimizing the eigenvalue based objective function. To make a fair comparison among the
40
41 algorithms used, same number of generations (iterations) and same boundary limits are considered. Moreover, the
42
43 stopping criterion for all algorithms is also set at a maximum number of generations. The optimizer will stop if it
44
45 reaches the same fitness value for the maximum number of generation on a consecutive manner. Each algorithm is
46
47 run for 30 times and the average running time is also reported.
48
49
50 Table 7 Fitness value and time required by PSO, DE and IWO algorithms (30 runs)
51
52
53
54
55 It is found from Table 7 that the best fitness value achieved by DE (-1539.1) is better than those obtained by IWO (-
56
57 1533.9) or PSO (-1531.6). Moreover, the average fitness value obtained by DE (-1538.4) is far better than PSO (-
58
59 1520.5)and IWO (-1525.5) for the 30 runs. This indicates the superior ability of DE compared to IWO and PSO to
60
61
62 21
63
64
65
1
2
3
4 reach much closer to the optimal solution for all the runs conducted. More interestingly, it is discovered that even
5
6 the worst fitness achieved by the DE (-1537.2) is better than the bests achieved by either PSO (-1531.6) or IWO (-
7
8 1533.9). With respect to the computational time, the average achieved by the DE algorithm is little inferior to PSO
9
10 but better than IWO. However, as stated earlier, the fitness achieved by PSO within shorter computational time is
11
12 inferior to the one achieved by the DE. Hence, from overall observations of this Table it can be concluded that on an
13
14 average, the performance attained by DE is better than those obtained by PSO and IWO.
15
16
17 To further investigate the performance of the three algorithms, non-parametric statistical analyses of the data
18
19 obtained from 30 independent test runs are performed using SPSS software. At first, the one sample Kolmogorov-
20
21 Smirnov (KS) test is conducted where the null hypothesis, H0, assumes that the data sample fits normal distribution
22
23 while the alternative hypothesis, H1, assumes that the data sample does not fit the normal distribution with a
24
25 significance level of 0.05. From the analyses presented in Table 8, it is seen that none of the data sample is missed
26
27 by the test as in all cases N = 30 samples are considered. The mean of the data sample shows that the DE algorithm
28
29
outperforms the PSO and IWO in this regard. The value of standard deviation illustrates that the data samples for
30
31
DE algorithm (0.53047) are far more adjacent to the best solution as compared to PSO (5.65277) and IWO
32
33
(3.91369),which reflects greater stability. The p-value (Asymp. Sig. 2-tailed) for DE, PSO and IWO algorithm are
34
35
36 found as 0.647, 0.711 and 0.942, respectively, which indicates that the data samples obtained from 30 independent
37
38 runs by each of the three algorithms do not fit normal distribution due to the rejection ofH0 and acceptance ofH1.
39
40
41 Table 8 One sample Kolmogorov-Smirnov test results
42
43 Table 9 Paired sample t- test results
44
45
46
47
48 Next, the paired t-test is run with the 30 samples to evaluate each algorithm distinctly and to discover further
49
50 differences among them. The results for this test are presented in Table 9. The H0 in this case assumes that the
51
52 means of the data sets are equal and H1 assumes the alternative statement that the means are unequal with a
53
54 significance level of 0.05. As observed from Table 9, all the paired p-values (Sig. 2-tailed) are less than 0.05 and
55
56 thus show a significant level to reject H0. So, this study shows 95% confidence level that the three pairs are
57
58 significantly different from each other. Moreover, the correlation obtained for the pairs involving DEare much
59
60
61
62 22
63
64
65
1
2
3
4 smaller than the PSO-IWO pair. Thus, from this test it can be concluded that the DE algorithm behaves differently
5
6 in a statistical manner compared to other two and gives better performance.
7
8
9
10
11
12 VI. Conclusion
13
14
15 A simple PI controller based approach is proposed for improving the small signal dynamic response of a grid
16
17 connected SOFC plant where the controller parameters are tuned by the DE algorithm. Performance of DE is
18
19 compared with those of PSO and IWO from different aspects. Eigenvalue based optimization is performed for
20
21 optimal tuning of the PI controller parameters. The superiority of eigenvalue based objective function over time
22
23 domain based objective function lies in the fact that the computation time can be saved in the first instance. The
24
25 results obtained by eigenvalue analysis does not clearly indicate the superior performance of DE compared to PSO
26
27 and IWO as only the best outcome of 30 different runs are considered for this part of analysis. To investigate the
28
29 overall performance of three algorithms, computational time and mean fitness value is also taken into consideration.
30
31 It is found that, even the lowest mean of the fitness function achieved by DE is superior to those obtained by the
32
33 other two. Although the run time result shows that the PSO can reach to a minimum little faster than DE and much
34
35 faster than IWO on an average, due to the inferiority of the minimum value of the fitness function obtained, it fails
36
37 to be regarded as the best optimizer. Again, the parameters obtained by PSO hit either the upper or the lower limit of
38
39 the boundary on most number of occasions whereas for DE, no such incidents are encountered. Three different types
40
41 of load variations are considered for the time domain based results. It is found that the optimal parameters obtained
42
43 by the proposed eigenvalue based objective function can successfully track all of these variations. The non-
44
45
parametric statistical test results show that the data sample obtained by these algorithms for 30 independent runs are
46
47
not correlated and the smallest standard deviation is recorded for DE. In conclusion, considering all the
48
49
aforementioned discussions, it can be said that the overall performance of DE is better than those of PSO and IWO
50
51
for the study under consideration.
52
53
54
Conflict of Interests
55
56
57 The authors declare that they have no conflict of interest.
58
59
60 References:
61
62 23
63
64
65
1
2
3
4 Pepermans, G., J. Driesen, D. Haeseldonckx, R. Belmans and W. D’haeseleer (2005). "Distributed
5
6
generation: definition, benefits and issues." Energy policy33(6): 787-798.
7 Choudhury, A., H. Chandra and A. Arora (2013). "Application of solid oxide fuel cell technology for power
8 generation—A review." Renewable and Sustainable Energy Reviews20: 430-442.
9 Jain, S., J. Jiang, X. Huang and S. Stevandic (2012). "Modeling of fuel-cell-based power supply system for
10 grid interface." Industry Applications, IEEE Transactions on48(4): 1142-1153.
11
Du, W., H. Wang and H. Cai (2013). "Modelling a grid‐connected SOFC power plant into power systems
12
13 for small‐signal stability analysis and control." International Transactions on Electrical Energy
14 Systems23(3): 330-341.
15 Gelen, A. and T. Yalcinoz (2015). "The modeling and simulation of thermal based modified solid oxide
16 fuel cell (SOFC) for grid-connected systems." Acta Scientiarum. Technology37(2): 211-219.
17 Bhuyan, K. C., S. Padhee, R. K. Patjoshi and K. Mahapatra (2014). "Hysteresis Control of Power
18
19 Conditioning Unit for Fuel Cell Distributed Generation System." World Academy of Science, Engineering
20 and Technology, International Journal of Electrical, Computer, Energetic, Electronic and Communication
21 Engineering8(7): 1166-1171.
22 Sedghisigarchi, K. and A. Feliachi (2006). "Impact of fuel cells on load-frequency control in power
23 distribution systems." Energy Conversion, IEEE Transactions on21(1): 250-256.
24
Nayeripour, M. and M. Hoseintabar (2013). "A new control strategy of solid oxide fuel cell based on
25
26 coordination between hydrogen fuel flow rate and utilization factor." Renewable and Sustainable Energy
27 Reviews27: 505-514.
28 Saha, A., S. Chowdhury, S. Chowdhury and Y. Song (2007). "Application of solid-oxide fuel cell in
29 distributed power generation." IET Renewable Power Generation1(4): 193-202.
30 Knyazkin, V., L. Söder and C. Canizares (2003). Control challenges of fuel cell-driven distributed
31
32 generation. Power Tech Conference Proceedings, 2003 IEEE Bologna, IEEE.
33 Li, Y., S. Rajakaruna and S. Choi (2007). "Control of a solid oxide fuel cell power plant in a grid-connected
34 system." Energy Conversion, IEEE Transactions on22(2): 405-413.
35 Li, Y., Q. Wu and H. Zhu (2015). "Hierarchical Load Tracking Control of a Grid-Connected Solid Oxide Fuel
36 Cell for Maximum Electrical Efficiency Operation." Energies8(3): 1896-1916.
37
38
Chatterjee, K., R. Shankar and A. Kumar (2014). "Fuzzy Logic Based Controller for a Grid-Connected Solid
39 Oxide Fuel Cell Power Plant." Journal of fuel cell science and technology11(5): 051005.
40 Hajizadeh, A. and M. Aliakbar-Golkar (2007). "Fuzzy control of fuel cell distributed generation systems."
41 Iranian Journal of Electrical and Electronic Engineering3(1): 31-41.
42 Mammar, K. and A. Chaker (2009). "Fuzzy logic control of fuel cell system for residential power
43
generation." Journal, of ELECTRICAL ENGINEERING60(6): 328-334.
44
45 Bhuyan, K. C. and K. Mahapatra (2011). An intelligent control of solid oxide fuel cell voltage. Power and
46 Energy Systems (ICPS), 2011 International Conference on, IEEE.
47 Sudheer, K. and P. Prasanna (2011). "Fuzzy and PI Control of Hybrid Fuel Cell/Battery Distributed
48 Generation Systems." International Journal of Modeling and Optimization1(2): 95.
49 Albertos, P. and A. Sala (1998). "Fuzzy logic controllers, advantages and drawbacks." IEEE transactions
50
51 on control system technology.
52 Taher, S. A. and S. Mansouri (2014). "Optimal PI controller design for active power in grid-connected
53 SOFC DG system." International Journal of Electrical Power & Energy Systems60: 268-274.
54 Kundur, P., N. J. Balu and M. G. Lauby (1994). Power system stability and control, McGraw-hill New York.
55 Huang, B., Y. Qi and A. M. Murshed (2013). Dynamic modeling and predictive control in solid oxide fuel
56
cells: first principle and data-based approaches, John Wiley & Sons.
57
58 Du, W., H. Wang, X. Zhang and L. Xiao (2012). "Effect of grid-connected solid oxide fuel cell power
59 generation on power systems small-signal stability." IET Renewable Power Generation6(1): 24-37.
60 Nise, N. S. (2011). "Control Systems Engineering, John Wiley&Sons." Inc, New York.
61
62 24
63
64
65
1
2
3
4 Vesterstrøm, J. and R. Thomsen (2004). A comparative study of differential evolution, particle swarm
5
6
optimization, and evolutionary algorithms on numerical benchmark problems. Evolutionary
7 Computation, 2004. CEC2004. Congress on, IEEE.
8 El Ela, A. A., M. Abido and S. Spea (2009). "Optimal power flow using differential evolution algorithm."
9 Electrical Engineering91(2): 69-78.
10 Qin, A. K., V. L. Huang and P. N. Suganthan (2009). "Differential evolution algorithm with strategy
11
adaptation for global numerical optimization." Evolutionary Computation, IEEE Transactions on13(2):
12
13 398-417.
14 Wang, S.-K., J.-P. Chiou and C.-W. Liu (2009). "Parameters tuning of power system stabilizers using
15 improved ant direction hybrid differential evolution." International Journal of Electrical Power & Energy
16 Systems31(1): 34-42.
17 Campanari, S. and P. Iora (2004). "Definition and sensitivity analysis of a finite volume SOFC model for a
18
19 tubular cell geometry." Journal of Power Sources132(1): 113-126.
20 Sedghisigarchi, K. and A. Feliachi (2004). "Dynamic and transient analysis of power distribution systems
21 with fuel Cells-part I: fuel-cell dynamic model." Energy Conversion, IEEE Transactions on19(2): 423-
22 428.Murshed, A. M., B. Huang and K. Nandakumar (2007). "Control relevant modeling of planer solid
23 oxide fuel cell system." Journal of Power Sources163(2): 830-845.
24
Felder, R. M. and R. W. Rousseau (2008). ELEMENTRY PRINCIPLES OF CHEMICAL PROCESSES, (With CD),
25
26 John Wiley & Sons.
27
28
29
30 Appendix:
31
32
33 SOFC data:
34
35
E0 =1.28 V R = 8.3146 J/mol-°C F = 96487 C/ mol N0 = 384
36
37
38
T0 = 923° C Tref = 25° C T2 = 25° C ΔĤr0 = -2.4183kJ/mol
39
40
41
42
Kh2 = 8.43e-4 kmol/(atm-s) Ko2 = 2.52 e-3kmol/(atm-s) Kh2O = 2.81 e-4 kmol/(atm-  H0 =26.1sec
s) 2
43
44
45  O0 = 2.91 sec  h0 O = 78.3 sec me = 1.1 kg C p =1e4
46 2 2

47
48
49 r0 = 0.126 ohms α = -2870 Tin = 900° C
50
51
52
53
54
55 Grid data:
56
57
58
Egrid= 1.0 pu Rf= 0 Lf = 0.7163 pu ω0 = 1.0 pu
59
60
61
62 25
63
64
65
1
2
3
4
Initial operating data:
5
6
7
8 nh2in = 1.2 mol/s nO2in = 0.6468 mol/s Ifcref = 193.35 A Ifcpu = 0.6445pu
9
10 Pfcpu = 0.6983pu Qgrid = 0 Kd = 0.6 Kq = 0.4
11
12 dc = 0.3499 Vdc = 1.6667 pu id = 0.6983pu iq = 0
13
14
15 Ph2 = 0.9671atm PO2 = 0.1803atm Ph2O = 1.3692atm
16
17
18
19
20
Base Values:
21
22
23
24 Ifcb = 300 A Pfcb = 97.33 kW Sbase = Pfcb Vfcbase = 324.46 V
25
26
27
28
29 Linearizing constants:
30
31
32
T24 T3 T2 T4 T3 T2
33 H1  dH 2  cH 2 2  bH 2 2  aH 2T2  (dH 2  cH 2  bH 2  aH 2T )
34 4 3 2 4 3 2
35
36
37
T24 T3 T2 T4 T3 T2
38 O1  dO2  cO2 2  bO2 2  aO2T2  (dO2  cO2  bO2  aO2T )
39 4 3 2 4 3 2
40
41
42 T24 T23 T22 T4 T3 T2
43 W1  dH 2O  cH 2O  bH 2O  aH 2OT2  (dH 2O  cH 2O  bH 2O  aH 2OT )
44 4 3 2 4 3 2
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 26
63
64
65
1
2
3
4
T24 T3 T2 T4 T3 T2
5 N1  dN 2  cN 2 2  bN 2 2  aN 2T2  (dN 2  cN 2  bN 2  aN 2T )
6 4 3 2 4 3 2
7
T4 T3 T2
8 N12  dN 2  cN 2  bN 2  aN 2T
9 4 3 2
10
T4 T3 T2
11 H12  dH 2  cH 2  bH 2  aH 2T
12 4 3 2
13
T4 T3 T2
14 O12  dO2  cO2  bO2  aO2T
15 4 3 2
16
T4 T3 T2
17 W12  dH 2O  cH 2O  bH 2O  aH 2OT
18 4 3 2
19
20
21 I 0fc N 0 RT 0 I 0fc N 0 RT 0 I 0fc N 0 RT 0
22 G1h  G2O  G3w 
23 2000 FPh02 4000 FPO02 2000 FPh02O
24
25
26
M1  T2  T 0  Tph  M1K h2  H1  G1h  Z1
1
27 Z1 
28 C p me
29
30
Tpo  M1KO2  O1  G2O  Z1 Tpw  M1Kh2O W1  G3w  Z1
31
32
33
34
35     
1 1 
36   R
0 00.5
Ph2 PO2  I fc r e  0 T  
0 0 T 0

37 I fc  N 0 
0
ln  2.52e  
4

  2 F  02
38 Ph02O T 

Tpr   
39
40
41 1000
42
43
44
45
Tpt  nnout
2

N1  M1 nnout
2
N12  nhout
2
H12  nhout
2O

W12  nOout2 O12  Tpr  nhout
2
H1  nhout
2O

W1  nOout2 O1 Z1
46
47
48   RT 0 P 0 P 00.5  
49  N0  ln 2 0 2  2.52e 4T 0  1.2586 
h O
   
1 1 
50   2 F Ph2O  K Hˆ 0 I 0 r 0 e  T0 T 0  
  r r  fc
51 Tpi     Z1
52  1000 500 500 
53  
 
54  
55
56
57
T24 T3 T2 T4 T3 T2
58
59
H t 2  dH 2  cH 2 2  bH 2 2  aH 2T2  (dH 2 in  cH 2 in  bH 2 in  aH 2Tin )
4 3 2 4 3 2
60
61
62 27
63
64
65
1
2
3
4
T24 T3 T2 T4 T3 T2
5 Ot 2  dO2  cO2 2  bO2 2  aO2T2  (dO2 in  cO2 in  bO2 in  aO2Tin )
6 4 3 2 4 3 2
7
8
9 Tpnh  (Tin  Tref ) H t 2 Z1 Tpno  (Tin  Tref )Ot 2 Z1
10
11

  H
12
13
H y1  Kihy 
K phyT 0 
 K phy K h2 Ph02  nhin2 0  2K r I 0fc

  2 K phy K rT 0 
14 H y2 
T0 H0 2 K h2 T0 H0 2 K h2 T0 H0 2
y3
15

  
16
17 K phyT 0 K poxT 0  K pox KO2 PO02  nOin20  K r I 0fc
18 H y4  H y 5   K ihy Ox1  K iox  Ox 2 
19 K h2 T0 H0 2 T0 O02 KO2 T0 O02
20
  K pox K rT 0 
21
22 K poxT 0
Ox3  Ox 4 
23
24
KO2 T0 O02 KO2 T0 O02
25
26
27 K R  kd 0 K pid
28 Ox 5   Kiox I d 1  K iid   pid f  I d 2  0 K pid Id 3 
 L Lf
29  f 
30
31
32
Vdc 0 K pid K R 
33 Id 4  I d 5   Kiid I q1  0 K piq I q 2  K iiq   piq f 
34 Lf  L
35  f 
36
37
38 K piq kq 0 Vdc 0 K piq K pv  d c 0  1
I q3  Iq4  I q 5   K iiq V1  
39
40 Lf Lf C dc I fcb 
41
42
43 kd 0 K pv kq 0 K pv id 0 K pv
44 V2   V3   V4  Kiv V5  
45 Cdc Cdc Cdc
46
47
48 iq 0 K pv I fc 0 K pv
49 V6   V7   V8   Kiv
50
51
Cdc C dc I fcb 
52
53
System Matrices:
54
55
56
57
58
59
60
61
62 28
63
64
65
1
2
3
4
5
 T 0
 0 0 0 
K h2 Ph02  nhin2 0  2 I 0fc K r  
2 K rT 0
0 0 0 

 T0 H 2 K h2 T  K h2 T0 H0 2 
0
6 0 H2
 
7
8
 T 0

K P0
h2O h2O  2 I 0fc K r  
K rT 0 
 0 0 0 0 0 
9  T0 O02 K h2OT0 h02O K O2 T0 O02 
 
10
11  0 0
T 0

K P0
h2O h2O  2 I 0fc K r  2 K rT 0
0 0 0 
 T0 h02O K T 0
K h2OT0 h02O
12  h2O 0 h2O

13  Tph Tpo Tpw Tpt Tpi 0 0 0 
14 A I fc

15  0 0 0 0 0 0 0 
 Tel 
16  
17  Rf kd 0 
 0 0 0 0 0  0
18 Lf Lf 
19  
 Rf kq 0 
20  0 0 0 0 0 0  
21  Lf Lf 
 
22
23  0 0 0 0 
 dc 0  1 
kd 0

Kq
0 
24 
 C dc I fcb  Cdc Cdc 

25
26
27  T0 
T

28 T K  0 0 0 Tpnh 0 0 0 0 
29  0 h2 H 2 
30  T0 
31  0 0 Tpno 0 0 0 0 
 T0 K O2 O02 
32  
Vdc 0
B 0 
33
0 0 0 0 0 0
34  Lf 
35  
36  0 0 0 0 0 0
Vdc 0
0 
37  Lf
 
38  I fc 0 id 0 iq 0 
 0 0 0 0 0    
C dc I fcb 
39
Cdc Cdc 
40  
41
42
43 T
 1 
44 0 0 0 0
Tel
0 0 0 0 0 0 0 0
45  
46 0 0 0 0 0 0 0 0 H y5 0 0 0 0
47  
Bsys  0 0 0 0 0 0 0 0 0 Ox 5 0 0 0
48
49 0 0 0 0 0 0 0 0 0 0 Id 5 0 0
 
50 0 0 0 0 0 0 0 0 0 0 0 I q5 0
51 0 V7 
52  0 0 0 0 0 0 0 0 0 0 0
53
54
55
56
57
58
59
60
61
62 29
63
64
65
1
2
3
4
5
 T 0
 0 0 0 
K h2 Ph02  nhin2 0  2 I 0fc K r  
2 K rT 0
0 0 0
T0
0 0 0 0


 T0 H 2 K h2 T  K h2 T0 H0 2 T0 K h2 H0 2 
0
6 0 H2
 
7  T 0

K O2 P  nOin20  I 0fc K r
0
O2  
K rT 0 T0 
8  0 0 0 0 0 0 0 0 0 
 T0 O02 K O2 T0 O02 K O2 T0 O02 T0 K O2 O02 
9  
10  0 0
T 0

K P0
h2O h2O  2 I 0fc K r  2 K rT 0
0 0 0 0 0 0 0 0 
11  T0 h02O K T 0
K h2OT0 h02O 
 h2O 0 h2O

12  Tph Tpo Tpw Tpt Tpi 0 0 0 Tpnh Tpno 0 0 0 
13  
 0 1 
14 0 0 0  0 0 0 0 0 0 0 0
 Tel 
15  
16 Asys  0 0 0 0 0 
Rf
0
kd 0
0 0
Vdc 0
0 0 
17  Lf Lf Lf 
 
18  Rf kq 0 Vdc 0 
19  0 0 0 0 0 0  0 0 0 0 
 Lf Lf Lf 
20 
 dc 0  1 kd 0 kq 0 id 0 iq 0 I fc 0 
21  0 0 0 0    0 0 0    
22 

 Cdc I fcb  Cdc Cdc Cdc Cdc  Cdc I fcb  
23  H y1 0 0 H y2 H y3 0 0 0 H y4 0 0 0 0 
24  0 Ox1 0 Ox 2 Ox 3 0 0 0 0 Ox 4 0 0 0 
 
25  0 0 0 0 0 I d1 Id 2 Id 3 0 0 Id 4 0 0 
26  
27  0 0 0 0 0 I q1 Iq2 I q3 0 0 0 Iq4 0 
 0 0 0 0 V1 V2 V3 V4 0 0 V5 V6 V7 
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 30
63
64
65
Table Click here to access/download;Table;All tables paper 3.docx

Table 1 Open loop eigenvalues

Eigenvalue Associated state Participation factor (%)

-0.0171 Te 100

-0.0383 Ph2 100

-0.3436 PO2 100

-0.0128 Ph2 o 100

-0.2000 I fc 100

0 ± 1.128i Δid, Δiq 45, 45

- 6.9768×10-17 ΔVdc 100

Table 2 Detail of different disturbances

Type of disturbance Duration/Instants of Magnitude of disturbance


applications
Step 5.0 sec onwards I fc = 30 A, Qref = 0.2 pu
Pulse 5-25 sec I fc = 30 A, Qref = 0.2 pu
Staircase 5-25 sec I fc = 30 A, Qref = 0.2 pu
25-45 sec I fc = 40 A, Qref = 0.25 pu
45-65 sec I fc = 50 A, Qref = 0.3 pu
> 65 sec I fc = 0 A, Qref = 0 pu
Table 3 Parameters used for PSO, DE and IWO algorithm

Parameter PSO DE IWO


Maximum Population size 100 100 100
Maximum number of generations 500 500 500
Stopping criteria (Number of consecutive
iterations with same value of objective 30 30 30
function )
c1, cognitive acceleration coefficient 2 - -
c2, social acceleration coefficient 0.01 - -
MF, Mutation factor - 0.9 -
CR, Crossover probability - 0.2 -
Initial size of population - - 25
Maximum deviation, sdmax - - 5
Minimum deviation, sdmin - - 0
Nonlinear modulation index, n - - 2
Sigma_initial - - 1
Sigma_final - - 0.001

Table 4 Upper and lower bounds of the controller gains

Parameter Lower bound Upper bound


Kphy -9 -4
Kihy -5 -1
Kpox -7 -4
Kiox -5 -1
Kpid -1.2 -0.6
Kiid -1.2 -0.6
Kpiq -1.2 -0.6
Kiiq -0.5 -0.1
Kpv 5 15
Kiv 2 6

Table 5 Eigenvalues obtained by PSO, DE and IWO (best from 30 runs)

Method PSO DE IWO States associated


-0.2065 + 0.0531i -0.2640 + 0.0000i -0.2364 + 0.0000i Ph2, nh2in
-0.2065 - 0.0531i -0.1817 + 0.0000i -0.1923 + 0.0000i
-0.0171 + 0.0000i -0.0171 + 0.0000i -0.0171 + 0.0000i Tcell
-0.1943 + 0.0000i -0.2654 + 0.0000i -0.4159 + 0.0000i Po2
-0.7018 + 0.0000i -0.7616 + 0.0000i -0.7758 + 0.0000i no2in
Eigenvalues -1.5683 + 1.3134i -1.9344 + 1.1094i -1.4547 + 1.2381i id, kd
-1.5683 - 1.3134i -1.9344 - 1.1094i -1.4547 - 1.2381i
-0.1684 + 0.0000i -0.1042 + 0.0000i -0.1670 + 0.0000i iq
-0.8225 + 0.0000i -2.0831 + 0.0000i -2.8902 + 0.0000i dc
-0.5190 + 0.1750i -0.6889 + 0.0000i -0.2941 + 0.0256i Vdc, kq
-0.5190 - 0.1750i -0.4683 + 0.0000i -0.2941 - 0.0256i
-0.0128 + 0.0000i -0.0128 + 0.0000i -0.0128 + 0.0000i Ph2o
-0.2000 + 0.0000i -0.2000 + 0.0000i -0.2000 + 0.0000i Ifc

Table 6 Optimized parameters by PSO, DE and IWO algorithms (best from 30 runs)

Parameter PSO DE IWO


Kphy -8.2423 -8.9641 -8.5884
Kihy -1.0000 -1.0556 -1.0000
Kpox -4.0515 -5.0111 -6.2194
Kiox -1.0000 -1.4821 -2.3663
Kpid -1.1510 -1.1852 -1.1300
Kiid -0.7533 -0.6068 -0.7355
Kpiq -0.8848 -1.1907 -0.7931
Kiiq -0.1175 -0.1107 -0.1022
Kpv 5.0000 9.5125 9.9223
Kiv 2.0000 5.2806 2.0786

Table 7 Fitness value and time required by PSO, DE and IWO algorithms (30 runs)

Best result Average result Worst result


Parameter
PSO DE IWO PSO DE IWO PSO DE IWO
Elapsed time 3.0655 5.9904 11.866 8.8415 13.6697 21.448 56.2372 28.3484 62.6976
Best fitness
-1531.6 -1539.1 -1533.9 -1520.5 -1538.4 -1525.5 -1507.9 -1537.2 -1520.1
value

Table 8 One sample Kolmogorov-Smirnov test results

PSO_fitness DE_fitness IWO_fitness


N 30 30 30
Normal Parameters Mean -1520.5135 -1538.3946 -1525.5112
Std. Deviation 5.65277 .53047 3.91369
Most Extreme Absolute .128 .135 .097
Differences Positive .128 .135 .084
Negative -.087 -.092 -.097
Kolmogorov-Smirnov Z .700 .738 .529
Asymp. Sig. (2-tailed) .711 .647 .942

Table 9 Paired sample t- test results

Method Paired differences


Mean Correlation t df Sig. (2-tailed)
DE-PSO -17.8812 0.058 -17.344 29 0.000
DE-IWO -12.88344 -0.095 -17.645 29 0.000
PSO-IWO 4.99768 0.101 4.185 29 0.000
Nomenclature Click here to access/download;Table;nomenclature for ISTE
paper.docx

List of symbols:

E0 SOFC open Circuit Voltage


R Universal Gas Constant
T Fuel Cell Temperature
F Faraday’s Constant
N0 Number of cells in series
PH 2 Hydrogen partial pressure
PO2 Oxygen partial pressure
PH 2O Water vapor partial pressure
r0 Cell resistance for ohmic loss
 Constant coefficient
T0 Standard temperature
Ifc Cell output current
nHin2 Hydrogen inlet flow rate

nOin2 Oxygen inlet flow rate

nHin2O Water vapor outlet flow rate


Vdc DC capacitor voltage
Vfc Cell output voltage
dc Converter duty ratio
Cdc DC capacitor
id, iq dq component of grid current
Rf, Lf Filter resistance and inductance
kd, kq dq component of inverter switching signal
ed, eq dq component of grid voltage
Pgrid, Qgrid Grid active and reactive power
MF Mutation factor for DE
Figure Click here to access/download;Figure;Fig. 1.docx

Idc2 Idc1 Grid filter Grid utility


Hydrogen input nh2in Ifc va ea
dc-dc
dc-ac vb i a eb
SOFC Vfc conver Vdc Cdc
inverter vc i b
ter ec
Oxygen input no2in
ic Rf Lf

Switching
Duty ratio
pulses
dc
Kd Kq
Figure Click here to access/download;Figure;Fig. 2.docx

eb q

eq  0

θ ea

e  ed
d ω

ec
Figure Click here to access/download;Figure;Fig. 3.docx

0
Ph2 (atm)
Ph2r
-0.05
Ph2
-0.1
0 20 40 60 80 100 120
Time(sec)
0
PO2 (atm)

PO2r
-0.01
PO2
-0.02
0 20 40 60 80 100 120
Time(sec)
Ph2O (atm)

0.2

0.1

0
0 20 40 60 80 100 120
Time(sec)
5
Te ( C)
o

0
0 20 40 60 80 100 120
Time(sec)

40
I fcr
I fc (A)

20
I fc
0
0 20 40 60 80 100 120
Time(sec)
2
I d (p u)

I dr
1
I d
0
0 20 40 60 80 100 120
Time(sec)
2
I q (p u)

I
qr
0
I
q
-2
0 20 40 60 80 100 120
Time(sec)
2
Vdc (p u)

V
dcr
0
V
dc
-2
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 4.docx

Ph2 id
Ph 2 r  f 2 (I fcref ) n in
h2 idr  f3  I fcref  K d
K Kiid
K ph 2  ih 2 K pid 
s s

 PO 2 r  f1 (  I fcref ) iq
K
K p ox  iox
s iqr  f 4  Qgrid  K q
Kiiq
n in
O2 K piq 
s
 PO2

Vdc

Vdcr  f5  I fcref  d c
Kiv
K pv 
s
Figure Click here to access/download;Figure;Fig. 5.docx

0
PSO
-0.01 Ref
DE
-0.02 IWO

-0.03
(atm)

-0.04
h2
P

-0.05

-0.06

-0.07

-0.08
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 6.docx

0
Ref
-0.002 PSO
DE
-0.004
(atm)

-0.006
O2

-0.008
P

-0.01

-0.012

-0.014
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 7.docx

0.12

0.1
Ref
PSO
0.08 DE
IWO
I (p u)

0.06
d

0.04

0.02

0
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 8.docx

0
Ref
PSO
DE
-0.05
IWO

-0.1
I (p u)
q

-0.15

-0.2

-0.25
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 9.docx

0
Ref
-0.05 PSO
DE
-0.1 IWO
V (p u)

-0.15
dc

-0.2

-0.25

-0.3

-0.35
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 10.docx

0.01
DE
0 IWO
PSO
-0.01

-0.02
(atm)

-0.03
h2
P

-0.04

-0.05

-0.06

-0.07
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 11.docx

-3
x 10
2
DE
IWO
0
PSO

-2

-4
(atm)

-6
O2
P

-8

-10

-12

-14
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 12.docx

0.12
PSO
DE
0.1 IWO

0.08

0.06
I (p u)
d

0.04

0.02

-0.02
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 13.docx

0.1

0.05

PSO
-0.05
I (p u)

DE
IWO
q

-0.1

-0.15

-0.2

-0.25
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 14.docx

0.05

PSO
-0.05 DE
IWO
V (p u)

-0.1
dc

-0.15

-0.2

-0.25

-0.3
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 15.docx

0.02

-0.02 PSO
DE
(p u)

IWO
-0.04
h2
P

-0.06

-0.08

-0.1
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 16.docx

-3
x 10
2

0
PSO
-2 DE
IWO
-4
(p u)

-6
O2
P

-8

-10

-12

-14

-16
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 17.docx

0.16
PSO
0.14 DE
IWO
0.12

0.1
I (p u)

0.08
d

0.06

0.04

0.02

-0.02
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 18.docx

0.05
PSO
0 DE
IWO
-0.05

-0.1
I (p u)

-0.15
q

-0.2

-0.25

-0.3

-0.35
0 20 40 60 80 100 120
Time(sec)
Figure Click here to access/download;Figure;Fig. 19.docx

0.05
PSO
0 DE
IWO
-0.05

-0.1
V (p u)

-0.15
dc

-0.2

-0.25

-0.3

-0.35
0 20 40 60 80 100 120
Time(sec)
Author’s Response to Reviewers‘ Comments Click here to access/download;Author’s Response to Reviewers‘
Comments;Response to reviewers comments.docx

Response to Reviewer 1 comments:

Thank you for your valuable feedback concerning the readability of the paper. According to your
suggestion, we have thoroughly checked the manuscript and included commas wherever required. The
newly included commas are marked in red in the revised manuscript.

Response to Reviewer 2 comments:

Thanks a lot for your constructive feedback. According to your suggestion, we have incorporated a list of
nomenclature in tabular form in the revised submission.

S-ar putea să vă placă și