DSpace at VNU: Environment segregation of Er3+ emission in bulk sol-gel-derived SiO2-SnO2 glass ceramics

Electrical Power and Energy Systems 63 (2014) 1023–1029

Contents lists available at ScienceDirect

Electrical Power and Energy Systems
journal homepage: www.elsevier.com/locate/ijepes

Implementation of supervisory controller for solar PV microgrid system
using adaptive neural model
Ho Pham Huy Anh ⇑
FEEE, DCSELAB, Ho Chi Minh City University of Technology, VNU-HCM, Viet Nam

a r t i c l e

i n f o

Article history:
Received 20 July 2013
Received in revised form 21 June 2014
Accepted 23 June 2014
Available online 30 July 2014
Keywords:
Solar photovoltaics (solar PV)
Solar PV microgrid system
Back Propagation learning algorithm (BP)
Adaptive neural-based supervisory
controller
Modeling and identification

a b s t r a c t
This paper investigates a novel forward adaptive neural model which is applied for modeling and implementing of the supervisory controller of the solar PV microgrid system. The nonlinear features of the solar

PV microgrid system were thoroughly modeled based on the adaptive identification process using
experimental input–output training data. This paper proposes the novel use of a back propagation (BP)
algorithm to generate the adaptive neural-based supervisory controller for the solar PV microgrid system.
The simulation results show that the proposed adaptive neural-based supervisory controller trained by
Back Propagation learning algorithm yields outstanding performance and perfect accuracy.
Ó 2014 Elsevier Ltd. All rights reserved.

Introduction
Hybrid renewable energy systems can be classified into two
main types: grid-connected and standalone. The renewable energy
sources can be PV or wind generators (or both), according to the
availability of solar radiation or wind velocity (or both) at the system site. Batteries are often used as a backup source to supply the
system when the renewable energy source is unavailable. Other
backup sources can be used with or without batteries such as fuel
cells (e.g. electrolysers, supercapacitors and flywheel energy storage). Diesel generators could be used as secondary sources of
renewable energy. The standalone system might provide dc power,
ac power, or both dc and ac power [1–3]. The grid-connected systems can work on standalone mode when the utility grid is
unavailable. For the most part, fuel cells and diesel generators
are not used with such grid-connected systems. The supervisory
controllers manage the power according to the type and different
components of the system. The supervisory controllers could be
divided generally to two kinds; conventional-based and artificial
intelligence-based methods.
A small-scale hybrid PV-wind generation system with batteries
works only in standalone mode as proposed in [4]. The supervisory
controller with a fault ride through strategy is explained in Ref. [5].

⇑ Tel.: +84 08 39490415.
E-mail address:
/>0142-0615/Ó 2014 Elsevier Ltd. All rights reserved.

The supervisory controller of a hybrid wind-PV-fuel cell (FC)
energy system is proposed in [6–8]. Every source is connected
to the ac bus bar via an inverter to supply the load. The FC–
electrolyzer combination is used as a backup and long-term
storage system. The battery bank is used in the system as a shorttime backup to supply the transient power. At any given time, the
supervisory controller controls any excess wind-PV-generated
power to be supplied to the electrolyser. The hydrogen, which is
delivered to the hydrogen storage tanks by a gas compressor, is
consequently generated. If the generated power is less than the
load demand, the FC stack begins to produce energy for the load
using hydrogen from the storage tanks. A steady state model was
used in the papers with no dynamical results. This study demonstrates that the low voltage distribution network is supervised to
optimize energy flow and control power quality [9]. This kind of
system is supplied by renewable energy sources, diesel generators,
and energy storage backups. The system is controlled, according to
international power quality standards. The algorithm is universal
and adapts its control variables. A supervisory controller for the
hybrid PV-wind system with batteries is proposed in [10]. The PV
is directly connected in parallel with the batteries to supply the
ac load through a three phase inverter which is connected from
the other side to a wind generator. The power management strategy is simplified in this configuration as the batteries act as a constant voltage load line which charges both ways by the PV and the
wind generators. A dump load can be switched on with batteries
fully charged but the batteries are later disconnected to prevent


1024

H.P.H. Anh / Electrical Power and Energy Systems 63 (2014) 1023–1029

overcharging. One of the drawbacks is that there is no ability in
this scheme to provide PV or wind generators control. Furthermore, the batteries’ charging and discharging is not fully controlled. Recently, authors in [11,12] introduced the supervisory
control system for hybrid wind diesel microgrid.
Up to now, there are many researches focus on artificial intelligence-based methods applied to supervisory control of hybrid
microgrid systems. A standalone system with hybrid PV-diesel
power generators and flywheel backup energy storage system is
proposed in [13]. A pump is used as an auxiliary load to absorb
the extra power from the system. A fuzzy logic supervisory controller is proposed to manage the power from the generators to
the load. According to the generated PV power and the rotor speed
of the flywheel, the fuzzy controller adjusts the references for the
diesel generator output power and the pump demand. A fuzzy
logic supervisor is proposed also in Ref. [14] for a grid-connected
wind generated system. The supervisory controller regulates the
power of the wind generator according to the change in the grid
frequency. The pitch angle is controlled to match the reference
power generated by the supervisory controller. Hong et al. in
[15] developed of intelligent MPPT (maximum power point tracking) control for a grid-connected hybrid power generation system.
In a microgrid system [16], the PV generators could be used to
remove frequency deviations using fuzzy supervisory controller.
This controller increases or decreases the PV output power to
match a high frequency or a low frequency respectively. In Ref.
[17], the fuzzy supervisor controls the pitch angle of a fixed speed
wind generator and the reactive power output of the static VAR
compensator to smooth the wind generator output power and regulate the grid voltage respectively. A neural networks-based supervisory controller manages the power in a PV standalone system
with batteries. Two neural networks are used: one neural network
for direct control and the second to adapt the first one to optimize
the system’s operation [18]. Other applications of neural and fuzzy
techniques in supervisory control of microgrid systems were investigated in [19–22]. Artificial bee colony-based approach was
applied in [23] as to solve the problem of capacitor placement
for net saving maximization and system stability enhancement in

distribution networks. The drawback of these researches is that
the proposed intelligent supervisory controllers were unable to
adaptively generate the switching control outputs. Unfortunately,
up to now, the use of adaptive neural network-based supervisory
controller for the microgrid systems has not yet been adequately
studied.
To overcome this gap, this paper proposes the novel use of
adaptive neural MIMO model to generate the supervisory controller for the solar PV microgrid systems. The Back Propagation (BP)
learning algorithm is used to process the experimental input–output data that is measured from the optimal desired operation of
the solar PV microgrid systems as to optimize all nonlinear and
dynamic features of this system. Thus, the BP algorithm optimally
generates the appropriate neural weightings to perfectly characterize the features of the supervisory controller for the solar PV
microgrid systems. These good obtained results are due to proposed adaptive neural MIMO model combines the extraordinary
approximating capability of the neural system with the powerful
predictive and adaptive potentiality of the nonlinear ARX structure
that is implied in the proposed adaptive neural-based model. Consequently, the proposed method of the generation of the adaptive
supervisory controller for the solar PV microgrid systems has successfully modeled the nonlinear features of the desired operation
of the solar PV microgrid system with good performance.
The rest of the paper is organized as follows. Section ‘Implementation of supervisory control for the solar PV microgrid system’
introduces the implementation of supervisory controller in solar
PV microgrid systems. Section ‘Adaptive neural MIMO model for

supervisory control of the solar PV microgrid system’ presents
the novel adaptive neural MIMO model using for the implementation of supervisory controller in solar PV microgrid systems. The
results from the proposed adaptive neural-based supervisory controller are presented in Section ‘Identification and implementation
of the adaptive neural MIMO model for supervisory control of the
solar PV microgrid system’. Finally, Section ‘Conclusions’ contains
the concluding remarks.
Implementation of supervisory control for the solar PV
microgrid system

We consider an implementation of a supervisory controller for
the solar PV microgrid systems illustrated in Fig. 1. This scheme
introduces the novel use of adaptive neural MIMO model to
generate the supervisory controller for the solar PV microgrid
systems.
Fig. 1 illustrates the working principle of proposed supervisory
controller for the solar PV microgrid systems. The proposed neural
NARX-based supervisory controller is designed through two
phases: offline training phase and then online operating phase. In
the training phase, based on the experimental input–output data
measured from the optimal desired operation of the solar PV
microgrid system, the Back Propagation (BP) learning algorithm
is applied to optimally generate the appropriate neural weightings
which perfectly characterize the features of the supervisory controller for the solar PV microgrid systems. Then, in the operating
phase, the neural networks-based supervisory controller will
optimally manage the power in a PV grid-connected system. This
neural networks-based supervisory controller adapt well the input
variables including PV power available and required load power to
optimize the system’s operation via appropriate switching outputs
S1, S2, S3. This controller is concerned with the utility grid not with
controlling the local generators. The grid-connected systems can
work on standalone mode when the utility grid is unavailable
and in grid-connected systems, the utility grid is a secondary
source. Four modes of switching operation of the proposed neural-based supervisory controller for the solar PV microgrid systems
were tabulated in Table 1 as follows.
Adaptive neural MIMO model for supervisory control of the
solar PV microgrid system
The adaptive forward neural MIMO controller used in this paper
is a combination between the Multi-Layer Perceptron Neural Networks (MLPNN) structure and the Auto-Regressive with eXogenous
input (ARX) model. Due to this combination, adaptive forward neural MIMO model possesses both of powerful universal approximating feature from MLPNN structure and strong predictive feature

from nonlinear ARX model.
A fully connected 3-layer feed-forward MLP-network with n
inputs, q hidden units (also called ‘‘nodes’’ or ‘‘neurons’’), and m
outputs units is shown in Fig. 2.
In Fig. 2, w10,. . ., wq0 and W10,. . .,Wm0 are weighting values of
Bias neurons of Input Layer and Hidden Layer respectively.
Forwardly we consider an Auto-Regressive with eXogenous input
(ARX) model with noisy input, which can be described as

AðqÀ1 ÞyðtÞ ¼ BðqÀ1 Þuðt À TÞ þ CðqÀ1 ÞeðtÞ
with AðqÀ1 Þ ¼ 1 þ a1 qÀ1 þ a2 qÀ2
BðqÀ1 Þ ¼ b1 þ b2 qÀ1
CðqÀ1 Þ ¼ c1 þ c2 qÀ1 þ c3 qÀ2

ð1Þ


H.P.H. Anh / Electrical Power and Energy Systems 63 (2014) 1023–1029

1025

Fig. 1. Schematic of a supervisory controller for the solar PV microgrid systems.

Table 1
Four modes of switching operation of the proposed adaptive neural-based supervisory controller for the solar PV microgrid systems.
Mode

S1

S2

S3

Conditions

1
2
3
4

OFF
ON
ON
OFF

ON
ON
OFF
OFF

OFF
OFF
ON
ON

(Solar
(Solar
(Solar
(Solar

power)
power)
power)
power)

PW % 0
PW < PL (Consumed load power)
PW > PL (Consumed load power)
PW > PL (PL = 0)

where e(t) is the white noise sequence with zero mean and unit variance; u(t) and y(t) are input and output of system respectively; q is
the shift operator and T is the time delay.
From Eq. (2), not considering the noise component e(t), we have
the general form of the discrete ARX model in z-domain (with the
time delay T = nk = 1)
À1

À1

À2

Ànb

b1 z þ b2 z þ . . . þ bnb z
yðz Þ
¼
uðzÀ1 Þ 1 þ a1 zÀ1 þ a2 zÀ2 þ . . . þ ana zÀna

ð2Þ

in which na and nb are the order of output y(zÀ1) and input u(zÀ1)
respectively.
We investigate the potentiality of various simple adaptive neural MIMO models in order to exploit them in modeling, identification and control as well. The adaptive neural-based supervisory
controller of the solar PV microgrid system is investigated. Thus,
by embedding a 3-layer MLPNN (with number of neurons of hidden
layer equal 5) in a 1st order ARX model with its characteristic
equation induced from (3) as follows:
s1hat ðkÞ ¼ b11 pS ðkÞ þ b12 pL ðkÞ À a11 s1 ðk À 1Þ À a12 s2 ðk À 1Þ À a13 s3 ðk À 1Þ
s2hat ðkÞ ¼ b21 pS ðkÞ þ b22 pL ðkÞ À a21 s1 ðk À 1Þ À a22 s2 ðk À 1Þ À a23 s3 ðk À 1Þ
s3hat ðkÞ ¼ b31 pS ðkÞ þ b32 pL ðkÞ À a31 s1 ðk À 1Þ À a32 s2 ðk À 1Þ À a33 s3 ðk À 1Þ
ð3Þ

We will design the proposed adaptive neural–based supervisory
controller of the solar PV microgrid system (with na = 1, nb = 1,
nk = 1) with 5 inputs (including two input values pw(k), pl(k) and
three recurrent delayed output values s1(k À 1), s2(k À 1),

Fig. 2. Structure of feed-forward MLPNN.


1026

H.P.H. Anh / Electrical Power and Energy Systems 63 (2014) 1023–1029

s3(k À 1)) and three output values s1hat(k), s2hat(k) and s3hat(k). We
remember that two input values pw(k), pl(k), representing the
two power inputs [MW] of the solar PV and the load, respectively
and the three output values s1hat(k), s2hat(k) and s3hat(k) representing the responding switching output of the adaptive neural-based
supervisory controller. Its structure is shown in Fig. 3.
By this way, the fifteen parameters a11, a12, a13, b11, b12, a21, a22,

a23, b21, b22, a31, a32, a33, b31, b32 of the ARX structure of three
switching output variables s1hat(t), s2hat(t) and s3hat(t), respectively,
now become adaptively nonlinear and will be determined from the
weighting values Wij and wjl of the proposed adaptive neural
MIMO NARX model. This feature makes adaptive neural MIMO
NARX model very powerful in modeling, identification and in
model-based advanced control as well.
The prediction error approach, which is the strategy applied
here, is based on the introduction of a measure of closeness in
terms of a mean sum of square error (MSSE) criterion:

EN ðh; Z N Þ ¼

N
1 X
^ðt jhފT ½yðtÞ À y
^ðt jhފ
½yðtÞ À y
2N t¼1

ð4Þ

Based on the conventional error Back-Propagation (BP) training
algorithms, the weighting value is calculated as follows:

Wðk þ 1Þ ¼ WðkÞ À k

@EðWðkÞÞ
@WðkÞ

ð5Þ

with k is kth iterative step of calculation and k is learning rate which
is often chosen as a small constant value.
Concretely, the weights Wij and wjl of neural MIMO NARX are
then updated as:

W ij ðk þ 1Þ ¼ W ij ðkÞ þ DW ij ðk þ 1Þ

DW ij ðk þ 1Þ ¼ k Á di Á Oj
^i ð1 À y
^i Þðyi À y
^i Þ
di ¼ y

ð6Þ

with di is search direction value of ith neuron of output layer
(i = [1 ? m]); Oj is the output value of jth neuron of hidden layer
^i are truly real output and predicted output
(j = [1 ? q]); yi and y
of ith neuron of output layer (i = [1 ? m]), and

wjl ðk þ 1Þ ¼ wjl ðkÞ þ Dwjl ðk þ 1Þ

Dwjl ðk þ 1Þ ¼ k Á dj Á ul
m
X
dj ¼ Oj ð1 À Oj Þ di W ij

ð7Þ

Identification and implementation of the adaptive neural MIMO
model for supervisory control of the solar PV microgrid system
In general, the procedure which must be executed when
attempting to identify a dynamical system consists of four basic
steps.





STEP
STEP
STEP
STEP

1
2
3
4

(Getting Training Data).
(Select Model Structure).
(Estimate Model).
(Validate Model).

In Step 1, the identification procedure is based on experimental
input–output data values measured from the desired input–output
of the adaptive neural–based supervisory controller of the solar PV

microgrid system. The two input values pw(k), pl(k), representing
the two power inputs [MW] of the solar PV and the load and the
three desired referential output values s1hat(k), s2hat(k) and s3hat(k)
representing the responding switching output of the adaptive neural–based supervisory controller. Fig. 4a presents the collected
input–output data composes of the two input signals pw(k), pl(k)
applied to the neural–based supervisory controller of the solar
PV microgrid system and Fig. 4b introduces the referential output
values s1hat(k), s2hat(k) and s3hat(k).
Back Propagation (BP) learning algorithm based on the error
between the (s1, s2, s3) reference switching outputs and the
responding (s1hat, s2hat, s3hat) switching outputs of adaptive neural
MIMO NARX model to update the weights of proposed neural
MIMO NARX supervisory controller. Fig. 5 illustrates identification
scheme of the neural MIMO NARX supervisory controller using
proposed neural MIMO NARX model for solar PV microgrid system.
The second step relates to selecting the model structure. The
block diagram in Fig. 5 illustrates the identification scheme of
the proposed intelligent model. The proposed adaptive neural
MIMO NARX model structure was attempted. Its model structure
was presented in Fig. 3.
The third step estimates values for the trained adaptive Neural
NARX model. The optimal fitness value to use for the BP-based
optimization and identification process is calculated. The estimation result is presented in Fig. 6. This figure represents the fitness
convergence values of the proposed neural-based supervisory controller which correspond to adaptive neural NARX identified and
optimized with Back Propagation (BP) learning algorithm. The
fitness value of the proposed adaptive neural-based supervisory

i¼1

pl(k)

ps(k)

s3hat(k)
s3(k-1)

TWO POWER INPUT VALUES OF TRAINING DATA
50

POWER of SOLAR
PV [kW]

in which dj is search direction value of jth neuron of hidden layer
(j = [1 ? q]); Oj is the output value of jth neuron of hidden layer
(j = [1 ? q]) ; ul is input of lth neuron of input layer (l = [1 ? n]).

pl(k)

30
20
10
0
0

5

10

15

20

0

5

10

15

20

45

s2hat(k)

s2(k-1)
pl(k)

s1hat(k)
ps(k)
s1(k-1)

Fig. 3. Model structure of the adaptive neural-based supervisory controller of the
solar PV microgrid system.

POWER of
LOAD [kW]

ps(k)

40

40
35
30
25
20

time [hour]
Fig. 4a. Two power input signals pw(k), pl(k) of training data for identification
process.


1027

H.P.H. Anh / Electrical Power and Energy Systems 63 (2014) 1023–1029
THREE OUTPUT VALUES OF TRAINING DATA

S1

1
0.5
0
0

5

10

15

20

0

5

10

15

20

0

5

10

15

20

S2

1
0.5
0

S3

1
0.5
0

time [hour]
Fig. 4b. Three switching output signals of training data for identification process.

Fig. 5. Identification scheme of the neural-based supervisory controller using
proposed adaptive neural MIMO NARX model.

adaptive neural-based solar PV microgrid supervisory controller
are implied in the three responding output switching signals (s1,
s2, s3) from two power input values (pw(k), pl(k)).
The last step relates to validating the resulting nonlinear adaptive models. Applying the same training diagram in Fig. 5, a good
validating result demonstrates the performance of the resulting
forward neural MIMO NARX (FNMN) model presented in Fig. 7.
The error which is optimized nearly zero between the real reference output switching signals (s1, s2, s3) and the forward neural
MIMO NARX model responding output signals (sh1, sh2, sh3) asserts
the very good performance of proposed neural MIMO NARX
controller.
Furthermore, as for consolidating the performance of the
trained neural-based supervisory controller, we have applied
another set of daily input power Pw, Pl values (see Table 2) for testing the adaptive performance of the trained neural-based supervisory controller. The output switching S1, S2, S3 results precisely
illustrated in Fig. 8a and 8b once more confirms the efficiency of
the proposed adaptive neural-based supervisory controller for
the solar PV microgrid system.
Finally, Fig. 9 illustrates the auto-tuning variation of adaptive
ARX parameters of proposed forward neural MIMO NARX Model
of the adaptive neural-based supervisory controller. Concretely,

the fifteen parameters a11, a12, a13, b11, b12, a21, a22, a23, b21, b22
and a31, a32, a33, b31, b32 of the two 1st order ARX structure integrated in proposed FNMN11 model were adaptively auto-tuning
as illustrated in Fig. 9. These results show that the parameters of
the ARX structure integrated in proposed FNMN models now
become adaptively nonlinear and will be adaptively determined
from the optimized weighting values Wij and wjl of the forward
neural MIMO NARX model. This feature once more proves that
the proposed adaptive forward neural MIMO NARX (FNMN) supervisory controller is very powerful and adaptive in identification
and in model-based advanced control as well.
Table 3 tabulates the optimized weighting values of the proposed forward neural MIMO NARX model. The final structures of
proposed neural MIMO NARX model identified and optimized by
BP learning algorithm are shown in Fig. 3.

FITNESS CONVERGENCE OF ADAPTIVE NEURAL MIMO NARX MODEL IDENTIFICATION

0

10

VALIDATION OF ADAPTIVE NEURAL MIMO NARX MODEL IDENTIFICATION

1

-2

S1

10
-4

0

-6

2
0
-2

10

0
10

error

ERROR

S1 ref
S1 neural model

0.5

-8

x 10

0

10

5

10

15

20

5

10

15

20

-3

1

S2

-10

10

S2 ref
S2 neural model

0.5

0

200

300

400

500

600

700

800

900

ITERATIONS
Fig. 6. Fitness convergence of proposed adaptive neural-based supervisory controller identification.

controller identification produces an excellent global optimal value
(equal to 0.00000000086).
These good results are due to how the proposed model combines the extraordinary approximating capability of the neural system with the powerful predictive and adaptive potentiality of the
nonlinear NARX structure that is implied in the adaptive neural
MIMO NARX model. Consequently, the complex features of the

0

5

10

15

0

5

10

15

20

1
0.5
0
-0.5

1

S3

100

error

0

error

-12

10

20
S3 ref
S3 neural model

0.5
0
0

5

10

15

20

0

5

10

15

20

0.5
0
-0.5
-1

time [hour]
Fig. 7. Validation of proposed adaptive neural-based supervisory controller
identification.


1028

H.P.H. Anh / Electrical Power and Energy Systems 63 (2014) 1023–1029

Table 2
Set of daily power Pw, Pl input values for testing the adaptive performance of the
trained neural-based supervisory controller.

0
1
2
3
4
5
6
7
8
9

10
11
12
13
14
15
16
17
18
19
20
21
22
23

0
0
0
0
0
0
0
20
73
80
154
179
187
179
160

104
38
0
0
0
0
0
0
0

33
37
39
40
43
58
60
70
75
68
60
50
62
52
54
45
40
62
80
73

60
48
40
32

À33
À37
À39
À40
À43
À58
À60
À50
À2
12
94
129
125
146
106
59
À2
À62
À80
À73
À60
À48
À40
À32

S1

Putility (MW)

ON ? 1, OFF ? 0
S1

S2

S3

0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
0
0

0
0
0
0
0

1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1

0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0

0

POWER of SOLAR
PV [MW]

S2 ref
S2 neural model

5

10

15

20

25

S3 ref
S3 neural model

0.5

0

5

10

15

20

25

Time [hour]

Fig. 8b. Consolidation testing of proposed adaptive neural-based supervisory
controller.

Adaptive NARX parameters' auto-tuning of proposed neural MIMO NARX model
20
10
0
-10
-20
-30
-40
-50
-60
-70
-80

0

POWER of
LOAD [MW]

25

0

50

20

20

0.5

0

100

15

15

1

150

10

10

0

200

5

5

1

TWO POWER INPUT VALUES OF CONSOLIDATION TESTING

0

S1 ref
S1 neural model

0.5
0

S2

PL (MW)

S3

PW (MW)

Adaptive NARX parameters' Values

t (hour)

THREE SWITCHING OUTPUT VALUES of CONSOLIDATION TESTING
1

0

5

25

10

15

20

25

30

35

40

45

time (samples X 30minutes)

80

Fig. 9. Adaptive NARX parameters’ auto-tuning of proposed neural MIMO NARX
supervisory controller.

60
40
20

0

5

10

15

20

conventional supervisory controller using relays and digital logic
control circuits which required high hardware cost, maintaining
fee, unable with variable input values (of solar power and load
power) and other disadvantages. On the contrary, the novel adaptive neural MIMO NARX-based supervisory controller using adaptively switching soft-computing control which required low
software cost, maintaining free, highly adaptive performance with
variable input values (from solar power and load power) and other

25

Time [hour]
Fig. 8a. Two power input signals pw(k), pl(k) of data for consolidation testing.

In summary, for comparing between the results respectively
obtained using the novel neural-based controller and the conventional supervisory controller, it convincingly shows that the

Table 3
Optimized weights of forward neural MIMO-NARX controller – total number of weighting values = 68.
j

wji – weights of Input Layer

i

1
2
3
4
5
0

Wj0 –
weight of
Bias
Input
layer

Wkj –
weights
of
Hidden
layer

i

k=1

k=2

k=3

0.1669
0.0075
0.0251
À0.0032
À0.0254

2.5321
À13.865
41.918
À5.7831
À45.182

2.568
À51.631
3.3732
À31.643
À28.857

1

2

3

4

5

6

7

8

9

0

À0.003
À0.019
À0.004
À0.027
0.006

À1.0561
0.1474
À0.2834
0.0142
À0.3565

1.023
À0.096
0.311
À0.109
0.329

0.0816
0.0091
0.0196
0.0047

À0.019

À1.056
0.147
À0.283
0.0136
À0.356

1.023
À0.095
0.311
À0.109
0.329

À0.167
0.0027
0.02635
À0.0043
À0.0115

À0.699
0.0386
0.0182
0.0349
À0.017

0.492
À0.037
0.038
À0.016

À0.013

0.016
0.00338
0.00848
À0.0105
0.00306

Wk0 –
weight of
Bias
Hidden
layer

À5.134

Wkj –
weights
of
Hidden
layer

Wk0 –
weight of
Bias
Hidden
layer

À12.923

Wkj –
weights
of
Hidden
layer

Wk0 –
weight of
Bias
Hidden
layer

À6.1125


H.P.H. Anh / Electrical Power and Energy Systems 63 (2014) 1023–1029

advantages. Furthermore, in comparison with the other intelligent
supervisory controllers introduced in [16,17,19,22,23], the
proposed adaptive neural-based controller can online adaptively
generate the switching control outputs, which seems unable with
the other intelligent supervisory controllers previously suggested.
Conclusions
In this paper a new approach of forward neural MIMO NARX
model firstly utilized in modeling and identification of the adaptive
supervisory controller applied in the solar PV microgrid system.
Training and testing results showed that the newly proposed adaptive neural MIMO NARX model presented in this paper can be used
in online control with better dynamic property and strong robustness. This proposed intelligent neural MIMO NARX model is quite
suitable to be applied for the modeling, identification and control
of various hybrid PV-wind-fuel cell microgrid systems, including

linear and nonlinear processes without concerns of large change
in external environments.
Acknowledgement
This research is funded by Vietnam National Foundation for
Science and Technology Development (NAFOSTED) under Grant
Number 107.04-2012.23.
References
[1] Khan M. Supervisory hybrid control of a wind energy conversion and battery
storage system. Graduate Department of Electrical and Computer Engineering,
University of Toronto, PhD thesis; 2008.
[2] Lemmon M, He K, Markovesky I. Supervisory hybrid systems. IEEE Control Syst
Mag 1999;19(4):42–55.
[3] Antsaklis P, Nerode A. Guest editorial hybrid control systems an introductory
discussion to the special issue. IEEE Trans Autom Control 1998;43(4):457–60.
[4] Liu C, Chau K, Zhang X. An efficient wind–photovoltaic hybrid generation
system using doubly excited permanent-magnet brushless machine. IEEE
Trans Industr Electron 2010;57(3):831–9.
[5] Khan M, Iravani M. Hybrid control of a grid-interactive wind energy conversion
system. IEEE Trans Energy Convers 2008;23(3):895–902.
[6] Wang C, Nehrir M. Power management of a stand-alone wind/photovoltaic/
fuel cell energy system. IEEE Trans Energy Convers 2008;23(3):957–67.

1029

[7] Ipsakisa D, Voutetakisa S, Seferlis P, Stergiopoulosa F, Elmasides C. Power
management strategies for a stand-alone power system using renewable
energy sources and hydrogen storage. Int J Hydrogen Energy 2009;34(16):
7081–95.
[8] Gebre T, Undeland T, Ulleberg O, Vie P. Optimal load sharing strategy in a
hybrid power system based on PV/fuel cell/battery/supercapacitor. In: In:

International conference on clean electrical power. Itali: Capri; 2009. p. 141–6.
[9] Irena W et al. A power-quality management algorithm for low-voltage grids
with distributed resources. IEEE Trans Power Delivery 2008;23(2):1055–62.
[10] Daniel S, Gounden N. A novel hybrid isolated generating system based on PV
fed inverter assisted wind-driven induction generators. IEEE Trans Energy
Convers 2004;19(2):416–22.
[11] Rajasekaran M, Merabet A, Ibrahim H, Beguenane R, Thongam JS. Control
system simulation for stand-alone hybrid wind diesel system. In: Electrical
Power and Energy Conference (EPEC), 2013 IEEE, Halifax, NS; 2013.
[12] Rajasekaran M, Merabet A, Ibrahim H, Beguenane R, Thongam JS. Control
system for hybrid wind diesel based microgrid. In: Electrical Power and Energy
Conference (EPEC), 2013 IEEE, Halifax, NS; 2013.
[13] Krichen L, Ouali A, Boukettaya G. Fuzzy logic supervisor for power control of an
isolated hybrid energy production unit. Int J Electr Power Eng
2007;1(3):279–85.
[14] Mokadem EE, Courtecuisse V, Saudemont C. Fuzzy logic supervisor-based
primary frequency control experiments of a variable-speed wind generator.
IEEE Trans Power Syst 2009;24(1):231–43.
[15] Hong CM, Ou TC, Lu KH. Development of intelligent MPPT (maximum power
point tracking) control for a grid-connected hybrid power generation system.
Energy, Elsevier 2013;50:270–9.
[16] Datta M, Senjyu T, Yona A, Funabash T. A fuzzy control based coordinated
method for isolated power utility connected clustered photovoltaic systems to
provide frequency control. In: IEEE 15th international conference on
intelligent system applications to power systems, Curitiba, Brazil; 2009,
p. 1–6.
[17] Krichen L, Francois B, Ouali A. A fuzzy logic supervisor for active and reactive
power control of a fixed speed wind energy conversion system. Electric Power
Syst Res 2008;78(3):418–24.
[18] Welch R, Venayagamoorthy G. Optimal control of a photovoltaic solar energy

system with adaptive critics. in: IEEE international joint conference on neural
networks, vols. 1–6, Orlando, USA; 2007, p. 985–90.
[19] Merlin VL, Santos R. Artificial neural network based approach for antiislanding protection of distributed generators. Int J Control Automat Syst
(IJCAS) 2014:339–48.
[20] Chaâbane N. A hybrid ARFIMA and neural network model for electricity price
prediction. Int J Electr Power Energy Syst 2014;55:187–94.
[21] Shahaboddin S, Dalibor P, Anuar NB. Sensorless estimation of wind speed by
adaptive neuro-fuzzy methodology. Int J Electr Power Energy Syst
2014:123–32.
[22] Duehee L, Baldick R. Short-term wind power ensemble prediction based on
Gaussian processes and neural networks. IEEE Transact Smart Grid
2014;5(1):501–10.
[23] El-Fergany A, Abdelaziz AY. Capacitor placement for net saving maximization
and system stability enhancement in distribution networks using artificial bee
colony-based approach. Int J Electr Power Energy Syst 2014;54:235–43.