Copyright (c) 2013 IEEE. Personal use is permitted.

For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

1

AbstractThis paper describes and evaluates an adaptive
neuro-fuzzy inference system (ANFIS)-based energy
management system (EMS) of a grid-connected hybrid system. It
presents a wind turbine (WT) and photovoltaic (PV) solar panels
as primary energy sources, and an energy storage system (ESS)
based on hydrogen (fuel cell FC-, hydrogen tank and
electrolyzer) and battery. All the energy sources use DC/DC
power converters in order to connect them to a central DC bus.
An ANFIS-based supervisory control system determines the
power that must be generated by/stored in the hydrogen and
battery, taking into account the power demanded by the grid, the
available power, the hydrogen tank level and the state-of-charge
(SOC) of the battery. Furthermore, an ANFIS-based control is
applied to the three-phase inverter, which connects the hybrid
system to grid. Otherwise, this new EMS is compared with a
classical EMS composed of state-based supervisory control
system based on states and inverter control system based on PI
controllers. Dynamic simulations demonstrate the right
performance of the ANFIS-based EMS for the hybrid system
under study and the better performance with respect to the
classical EMS.

Index Terms ANFIS, energy management system, energy
storage system, hybrid system, renewable energies.
NOMENCLATURE
P Active power (W)
Q Reactive power (VAr)
P Extra power to be generated (W)
L Tank level (%)
e Wind turbine rotational speed (rpm)
Dc Duty cycle (-)
I Current (A)
V Voltage (V)

Manuscript received July 5, 2013; revised October 4, 2013; accepted
October 31, 2013. This work was supported by the Spanish Ministry of
Science and Innovation under Grant ENE2010-19744/ALT.
Copyright (c) 2013 IEEE. Personal use of this material is permitted.
However, permission to use this material for any other purposes must be
obtained from the IEEE by sending a request to pubs-permissions@ieee.org.
P. Garcia, C. A. Garca, L. M. Fernandez and F. Llorens are with the
Department of Electrical Engineering, Superior Polytechnic School of
Algeciras, University of Cadiz, 11202 Algeciras (Cadiz), Spain (e-mail:
pablo.garcia@uca.es; carlosandres.garcia@uca.es; luis.fernandez@uca.es;
francisco.llorens@uca.es).
F. Jurado is with the Department of Electrical Engineering, Superior
Polytechnic School of Linares, University of Jan, 23700 Linares (Jan),
Spain (e-mail: fjurado@ujaen.es).
Indexes
ANFIS-I ANFIS-based inverter control system
ANFIS-S ANFIS-based supervisory control system
av Available
bat Battery
bus DC bus
con DC/DC converter
d Direct component
dem Demanded
fc Fuel cell
grid Grid
H2 Hydrogen
inj Injected to the grid
inv Inverter
lz Electrolyzer
net Net value
PI PI controller
pv Photovoltaic
q Quadrature component
ref Reference value
rw Renewable energy (photovoltaic and wind)
STATES State-based supervisory control system
wt Wind turbine
I. INTRODUCTION
owadays, renewable energy sources play an important
role in electrical energy generation [1]-[3]. Power
systems are undergoing substantial changes as generation
moves from large-centralized conventional power plants to
small-decentralized renewable power plants [4]. In fact, the
use in these small power plants of hybrid renewable energy
systems (HRES), mainly based on renewable sources and
ESS, presents some advantages over larger power plants, such
as environmental-friendly, high power quality, uninterruptible
service, cost savings, on-site generation, expandability, etc.
[5].
A typical HRES configuration combines several renewable
energy sources, such as WT and PV panels, with ESS. The
renewable sources are used as primary energy sources, which
are generating whenever there is wind or solar radiation.
However, because the sun irradiance and the wind speed are
uncontrollable parameters, a support energy source is needed
to increase the degree of controllability and operability of the
HRES. Traditionally, this function is performed by an ESS
ANFIS-Based Control of a Grid-Connected
Hybrid System Integrating Renewable Energies,
Hydrogen and Batteries
Pablo Garca, Carlos Andrs Garca, Luis M. Fernndez, Member, IEEE, Francisco Llorens and
Francisco Jurado, Senior Member, IEEE
N
Copyright (c) 2013 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

2
such as battery and hydrogen system (FC, electrolyzer and
hydrogen storage tank) [6]. These ESS are considered as an
effective solution to balance the generation and demand [7],
[8], supporting the renewable energy deficit when necessary,
and storing the primary energy excess when possible.
All renewable sources and ESS are connected together to a
central DC bus by means of power converters. These
converters are designed to deliver energy from the energy
sources, ensuring stable, sustainable and reliable operation [9].
In grid-connected mode, DC/DC and AC/DC power
converters interface the renewable sources and EES to a DC
bus, while DC/AC voltage source inverters are in charge of
delivering the active and reactive power to the grid [10]-[12].
HRES present a two-level control architecture. In the
highest level, the supervisory control system is responsible for
determining the reference power that must be generated
by/stored in the ESS. The lowest level is in charge of
controlling the converters associated to the renewable sources
and EES, so that the energy sources work as required by the
supervisory control system [9]. Most of the studies about
HRES control architectures consider stand-alone operation in
isolated sites [6], [13], due to the advances in renewable
energy technologies and power electronic converters, which
are used to convert the unregulated power generated from the
renewable sources into useful power at the load end.
Classical and intelligent control techniques have been
applied to the control system of stand-alone HRES [14].
Classical control techniques (i.e., based on state machine
controllers or PI controllers) require exact mathematical model
of the system and are very sensitive to parameter variations
[15]. Intelligent control techniques such as artificial neural
networks, fuzzy logic or neuro-fuzzy are more efficient and
robust than classical techniques, since they do not require an
exact model of the system, and improve the dynamic
behaviour of the system. Among the neuro-fuzzy systems,
ANFIS is faster in convergence when compared to the other
neuro-fuzzy models.
In the existing literature, grid-connected HRESs are not as
common as stand-alone HRESs. Furthermore, no references
have been found concerning the application of ANFIS to the
supervisory control system. When used, ANFIS has been
applied to the grid-connected inverter control [16], [17]. A
comparison between ANFIS and PI controllers for controlling
a single-phase inverter was presented in [16]. The
implementation of an ANFIS controller for a three-phase
inverter was studied in [17], in which the variables used to
control the power delivered to grid were the DC bus voltage
and the reactive power.
This paper presents a new ANFIS-based EMS for a HRES
composed of renewable energies, hydrogen and battery, and
an ANFIS-based control for the three-phase inverter, which
connects the HRES to grid. The main novelties of this paper
are: 1) the application of ANFIS to the supervisory control
system of a grid-connected HRES in order to determine the
power that must be generated by/stored in the ESS (hydrogen
and battery), taking into account the power demanded by the
grid, the available power, the hydrogen tank level and the
battery SOC; and 2) the application of ANFIS to the three-
phase inverter in order to properly control the power delivery
to grid by using the active and reactive power as control
variables. It is organized as follows. Section II describes the
grid-connected hybrid system under study. The new
hierarchical control architecture based on ANFIS is detailed in
Section III. Section IV describes the classical EMS used to
evaluate the performance of the ANFIS-based EMS. The
hybrid system is evaluated in Section V by a long simulation
of one year to check the performance of the ANFIS-based
supervisory control, and a short simulation to test the
performance of the ANFIS-based inverter control. Finally,
Section VI establishes the conclusions derived from this work.
II. GRID-CONNECTED HYBRID SYSTEM
Fig. 1 shows the grid-connected hybrid system under study
in this work, which is composed of WT and PV panels
(renewable and primary energy sources) and hydrogen
subsystem and battery (ESS). All of them use DC/DC power
converters in order to connect them to a central DC bus. The
whole system is connected to grid by a three-phase inverter. In
this system, the renewable sources are generating whenever
there is wind or solar radiation. When possible, renewable
energy is stored in the battery and/or in form of hydrogen by
using the electrolyzer, and this stored energy is recovered to
support the renewable production when necessary.
The hybrid system consists of the following equipment:
A 1.5 kW Bornay 1500 WT.
Nine 0.180 kW Eoplly EP125M/72 PV panels, with a
total rated of 1.6 kW.
A 14.48 kWh battery composed of 12 units of BAE
Secura 6 PVS 660 lead-acid battery.
A 0.48 kW Heliocentris HG-60 proton exchange
membrane (PEM) hydrogen electrolyzer.
Three metal hydride canister HS 760 hydrogen tanks
with a total capacity of 2280 liters.
A 1.2 kW Heliocentris Nexa 1200 PEM FC system.
A dynamic model of this grid-connected hybrid system is
implemented in MATLAB/Simulink environment, as
described below.
A. Wind Turbine
The WT presents a rated power of 1.5 kW. It presents a
two-blades turbine coupled to a three-phase permanent magnet
synchronous generator (PMSG).
This WT is represented by a model with the following
subsystems: turbine and generation system. The turbine model
expresses the mechanical power extracted from the wind,
which is a function of the wind speed and the blade tip speed
ratio, as defined by the actuator disk theory.
The generation system is composed of three-phase PMSG,
AC/DC converter and DC/DC converter, which are
represented by models included in SimPowerSystems [18].
The electrical and mechanical parts of the PMSG are
represented by a second-order state-space model. The AC/DC
and DC/DC converters are modeled by average-value
equivalent models. The DC/DC power converter, which
connects the WT to the hybrid system DC bus, is controlled by
Copyright (c) 2013 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

3
a torque reference-based maximum power point tracking
(MPPT) control in order to extract the maximum available
power from the WT. This MPPT control maintains the
operating point of the WT on its maximum power coefficient
for any wind speeds in the below-rated wind speed region,
modifying the duty cycle of the WT DC/DC converter, which
produces a variation of its rotational speed.
Furthermore, the WT generation system incorporates a
braking resistor at the DC bus, in which the power excess with
above nominal winds is dissipated to assure WT rated power.
B. PV panels
The PV system presents nine 0.180 kW PV panels, with a
total rated of 1.6 kW. A single-diode model, which is
composed of a current source and a parallel diode
(representing the ideal PV cell) with two resistances (series
and parallel resistances), is used to represent each PV panel.
This model presents suitable accuracy [19], and the parameters
are easy to find in the commercial datasheets [20], which
makes it perfect for the simulation of PV devices with power
converters.
A DC/DC power converter controlled by a MPPT controller
adapts the PV output voltage to the DC bus voltage. The
MPPT controller generates the duty cycle of the PV converter
to move the PV voltage to the voltage that corresponds to the
maximum power point (MPP). In this work, the MPP voltage
is defined as proportional to the PV open-circuit voltage [21].
Thus, a fractional open-circuit voltage algorithm is used as
MPPT algorithm to generate the duty cycle of the controller
due to its simplicity.
C. Hydrogen subsystem
The hydrogen subsystem is composed of 0.48 kW PEM
electrolyzer, 2280 liters hydrogen tank, and 1.2 kW PEM FC.
PEM electrolyzer and PEM FC are highly efficient units that
are suitable for autonomous and distribution operation [22].
The PEM electrolyzer uses electrical energy to produce
hydrogen from water. It is modeled by a resistance, in which
the power needed to produce hydrogen is consumed.
Faradays law is used to calculate the hydrogen produced
depending on the electrical current in the resistance [23]. The
hydrogen produced by the electrolyzer is stored in the
hydrogen tank.
In the hydrogen tank model, the hydrogen available in the
tank is obtained from the rate of incoming and outgoing
hydrogen, taking into account the ideal gas equation [24].
The PEM FC uses the hydrogen available in the tank in
order to produce electrical energy. It is modeled by a reduced
model of the complete model detailed in [25]. The validity of
the reduced model was demonstrated in [26], where both, the
reduced and complete models, were compared, showing
similar responses, but with a considerable reduction of the
computational time for the first one. Several control strategies
for FC vehicles [27]-[30] were evaluated by using this reduced
model. In this reduced model, the FC voltage depends on the
cell voltage and number of cells [25], [31]. The cell voltage is
determined as a function of the current density, temperature,
and water, hydrogen and oxygen partial pressures. The
compressor, which controls the air flow in the cathode to keep
constant the oxygen excess ratio, is represented by a first order
system. Finally, ideal air cooler and humidifier are considered
in the FC model [26].
D. Batteries
The hybrid system uses a 14.48 kWh lead-acid battery. This
type of battery is usually the least expensive storage battery
for any application, while still providing good performance
and life characteristics [32].
This battery is modeled by a resistance in series with a
variable voltage source [33].
E. DC/DC Converters
Each energy source of the hybrid system provides a variable
voltage, which depends on the current demand, at different
ranges. Therefore, it uses a pulse-width modulated (PWM)

Fig. 1. Grid-connected hybrid system under study.
Copyright (c) 2013 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

4
DC/DC converter [34] to transfer the output power to the
central DC bus.
The output voltage of the WT rectifier is higher than the DC
bus voltage. Thus, the WT plus rectifier system is connected
to the DC bus by using a buck-type unidirectional converter.
The connection of PV, FC and electrolyzer to the DC bus is
performed by using boost-type unidirectional converters. The
PV and FC converters transfer power from the source to the
DC bus, since both sources terminal voltages are lower than
the DC bus voltage. The electrolyzer converter transfers power
from the DC bus to the electrolyzer, whose terminal voltage is
higher than the DC bus voltage.
The battery uses a bidirectional converter, which allows the
power flow from the battery to the DC bus (boost-type) and
vice versa (buck-type).
Each DC/DC converter is modeled by using the two-
quadrant chopper model included in SimPowerSystems [18].
In this model, the chopper is represented by a simplified
version of the converter containing an average-value
equivalent model. It uses controlled voltage and current
sources in order to model the power-electronics switches,
enabling larger sample times while preserving the average
voltage dynamics.
F. Inverter
A three-phase IGBT inverter connects the DC bus of the
hybrid system to grid. This inverter is represented by the
model developed in [18]. In this model, the snubber capacitor
is eliminated, so that only the snubber resistance is taken into
account. Furthermore, the forward voltages of the IGBTs and
diodes are considered null. The inverter is PWM to produce
the three-phase 50 Hz sinusoidal voltage. It uses hysteresis
switching and controls the active and reactive power, as will
be shown below.
III. ANFIS-BASED ENERGY MANAGEMENT SYSTEM OF THE
HYBRID SYSTEM
Fuzzy logic control (FLC) is a technique very used in
control systems based on microprocessor, since it does not
require an exact model of system and it is insensitive to
variations of the parameters and operating points. It is based
on rule basis and membership functions (MF), which are
usually obtained by trial and error method, and thus being a
very time consuming and error-prone process [16], [35].
Artificial neural networks (ANNs) are mathematical models
with learning and parallel data processing abilities, which use
computational neurons (non-linear cells) organized in layers
and connected to each other by weight factors. ANNs are also
used in microprocessor control systems due to their non-linear
and adaptive structure, generalization skills, and design
independence from system parameters. However, their
disadvantages are the lack of rules for defining the structure
(cells and layers) due to their black box nature, and the
instruction problem of network [16], [35].
FLC and ANNs are complementary technologies in the
design of intelligent control systems [16], [35]. Neuro-fuzzy
systems combine the inference ability of fuzzy logic like a
human and the learning and parallel data processing abilities
of ANNs. With these systems, the development time is
reduced and the accuracy of the fuzzy model is improved.
ANFIS is one of the most successful neuro-fuzzy systems
developed by Jang in 1993 [36], which applies neural learning
rules to identify and tune the parameters and structure of a
fuzzy inference system, based only on the available data. Its
main characteristics are [35]-[38]: 1) the implementation is
easy; 2) the learning is fast and accurate; 3) it has strong
generalization skills; 4) the fuzzy rules make easier its
understanding; and 5) it is easy to incorporate both linguistic
and numeric knowledge for problem solving.
A typical architecture of an ANFIS [35]-[38] is shown in
Fig. 2, in which a circle indicates a fixed node, whereas a
square indicates an adaptive node. It presents a multi-layered
feed-forward network, in which each layer has a particular
function on the input signals. In this simple example, ANFIS
has two inputs (x, y) and one output (f), with five layers. In the
first layer, the nodes (A
i
, B
i
) contains the MF assigned to each
input (x, y). In the second layer, the system identifies the two
rules (rule 1: x=A2 and y=B1; rule 2: x=A2 and y=B2). The
third layer calculates the normalized firing strength of each
rule (w
1
, w
2
). The fourth layer includes the linear functions,
which are functions of the input signals (f1=p1x+q1y+r1,
f2=p2x+q2y+r2), taking into account the normalized firing
strength of each rule (w
1
, w
2
) calculated in the precedent layer.
Finally, the fifth layer computes the overall output by
summing all incoming signals.
MF parameters must be tuned by the training data. In this
paper, the adaptive error back-propagation method was used to
update the weights in each epoch. To avoid over-fitting
problems during the estimation, the data set was randomly
split into two sets: a training set (70% of the data) and a
validation data set (30%) [35], including the testing data set
(15%) used to check the generalization capability of resulting
ANFIS, and the checking data set (15%) to control the
potential for the model overfitting the data. Both data sets
(testing and checking) were different from the training data
set. The final error is defined by the root-mean-square
difference between the current and desired outputs.
In this work, ANFIS was used in both the supervisory
control and the grid-connected inverter control, and it was
implemented by using the Fuzzy Logic Toolbox of MATLAB
[39].

Fig. 2. Typical architecture of ANFIS [35]-[38].
A. ANFIS-Based Supervisory Control System
The ANFIS-based supervisory control system is responsible
for determining the power that must be generated by/stored in
the hydrogen (FC/electrolyzer) and battery, taking into
Copyright (c) 2013 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

5
account the power demanded by the grid, the available power,
the hydrogen tank level and the battery SOC.
Fig. 3 shows the overall configuration of the ANFIS-based
supervisory control system. This ANFIS control system has
three inputs, the battery SOC, the hydrogen tank level and the
net power, Pnet (difference between the power demanded by
the grid, P
dem
, and the renewable power generated by the WT
and PV panels, P
rw
). It has one output, the battery reference
power. The hydrogen reference power (FC and electrolyzer) is
calculated as the difference between the net power and the
battery reference power. These powers are limited in order to
avoid an over charge or discharge of the battery or hydrogen
tank.
The demanded power is determined by the power dispatch
system from the reference powers required by the system
operator, which are defined taking into account the maximum
available power in the hybrid system. The maximum available
power (P
av
) is calculated as the sum of the power generated by
the renewable sources (WT and PV) and the maximum
available power in the hydrogen and battery. The maximum
available power in the hydrogen is the maximum power that
can generate the FC if the hydrogen tank is emptied in a day
from its current hydrogen level. Similarly, the maximum
available power in the battery is the maximum power that can
deliver the battery during a day to achieve a final SOC of
20%. The demanded power should be equal or smaller than
the maximum available power.
Fig. 4 shows the bell shaped MF used in this work. The
battery SOC has five MF: very small (VS), small (S), medium
(M), high (H) and very high (VH). The hydrogen tank level
presents three MF: small (S), medium (M) and high (H).
Finally, the net power uses four MF: negative big (NB),
negative (N), positive (P) and positive big (PB).
To define these MF, 1200 training and testing data were
selected from the results obtained by the three EMS based on
flowcharts presented in [40] when applied to the hybrid
system under study in this work. In fact, each flowchart was
simulated with random inputs of battery SOC, hydrogen tank
level and net power in order to obtain the battery power values
used as training and testing data. The battery SOC and
hydrogen tank level varied between 0 and 100 %. The limit
values of the net power were the sum of the maximum
electrolyzer power and the maximum battery charge power
(smallest value), and the sum of the maximum FC power and
the maximum battery discharge power (highest value).
The final error, defined as the root-mean-square difference
between the current and desired outputs of the battery power,
was 0.3018.

Fig. 4. MF of the ANFIS-based supervisory control system. a) Battery SOC;
b) hydrogen tank level; and c) net power.
B. ANFIS-Based Grid-Connected Inverter Control System
An ANFIS-based control system is applied to the three-
phase inverter to regulate the output power delivered to grid.
The control structure adopted in this paper to provide the
active and reactive power demanded by the grid is shown in
Fig. 5. It presents two ANFIS controllers, one for the active
power and the other for the reactive power.
The ANFIS controller of active power (ANFIS_P)
determines the direct component of the current to be provided
by the inverter. It uses as inputs the error between the active
power demanded by the grid (P
grid
ref
) and the actual active
power injected by the inverter into the grid (P
grid
), and the
derivate of this error, interpreted as a prediction of future
errors, which helps to improve the system stability. Similarly,
the ANFIS controller of reactive power (ANFIS_Q) defines
the quadrature component of the current to be provided by the
inverter from the error and the derivative of the error between
the reactive power demanded by the grid (Qgrid
ref
) and the
actual reactive power injected by the inverter into the grid
(Q
grid
). The current components are transformed from the dq0
reference frame to the abc reference frame, which are used to
define the inverter switching pulses.

Fig. 3. ANFIS-based supervisory control system.
Copyright (c) 2013 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

6
To define the ANFIS controller, 1400 training and testing
data were selected from the simulations of the hybrid system
under study. In this case, the inverter was simulated with
random values in the duty cycle of the switches in order to
obtain random values for the current and output power of the
inverter. The collected data were used to implement the
ANFIS controller.

Fig. 5. ANFIS-based inverter control system.
Once again, the adaptive error back-propagation method
was used during the training. The final error, defined as the
root-mean-square difference between the current and desired
outputs of the direct (quadrature) component of the inverter
current for the ANFIS controller of active (reactive) power,
was 0.0607 (0.0178).
C. DC-DC Power Converters Control
The appropriate control of the DC-DC power converters
allows the battery, electrolyzer and FC to adjust the power to
the reference value defined by the supervisory control system.
Fig. 6 shows the control loops applied to the DC-DC power
converters.

Fig. 6. Control loops applied to the DC-DC power converters: a) FC
converter; b) electrolyzer converter; and c) battery converter.
In the case of the FC and electrolyzer, the reference current
of their DC-DC power converters is obtained by dividing the
reference power by the DC bus voltage. The duty cycle of
these converters is defined from a PI controller for the current
control loop.
The battery converter allows controlling the battery power,
and the DC bus voltage to the rated value of 24 Vdc. The
battery reference power defined by the supervisory control
system is modified by the output of a PI controller for the DC
bus voltage control loop. Therefore, during a change in the
demanded load or in the power generated by the different
energy sources, this controller produces an extra power to be
generated by the battery (Pbat), which allows controlling the
DC bus voltage. The reference current of the battery converter
is calculated by dividing the sum of this extra battery power
and the battery reference power by the DC bus voltage.
Finally, a PI controller for the current control loop is used to
determine the duty cycle of the battery converter.
IV. CLASSICAL ENERGY MANAGEMENT SYSTEM OF THE
HYBRID SYSTEM
In this section, the classical EMS used to test the
performance of the proposed ANFIS control system is
described.
The classical EMS is composed of state-based supervisory
control system based on states and inverter control system
based on PI controllers. The supervisory control system uses
the control scheme shown in Fig. 7 in order to determine the
power generated by/stored in the hydrogen and battery, taking
into account the power demanded by the grid, the available
power, the hydrogen tank level and the battery SOC. The
inverter control system presents two PI-controllers for
controlling the active and the reactive power demanded by the
grid. The control scheme is similar to that shown in Fig. 5, but
replacing the ANFIS controller and derivative term by a PI
controller. Otherwise, the control scheme implemented in the
DC-DC power converters is the same as described in Section
IV.C and shown in Fig. 6.

Fig. 7. Scheme of the supervisory control system based on states.
V. SIMULATION RESULTS
Two simulations were performed in order to test the HRES
controlled by the proposal ANFIS controllers by comparing it
with the HRES controlled by the classical EMS. The first one,
performed by the simulation of the hybrid system operation
during one year, was used to check the right performance of
the ANFIS-based supervisory control system in order to
satisfy the power demanded by the grid and control the battery
SOC, hydrogen level and battery, FC and electrolyzer power.
Furthermore, this simulation was used to compare the ANFIS-
Copyright (c) 2013 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

7
based supervisory control system with the classical
supervisory control system based on states. The parameters
used in the comparison were the battery efficiency (qbat),
hydrogen system efficiency (q
H2
), hybrid system efficiency
(qHS) and energy injected into the grid (Einv
inj
). The parameters
were calculated as follows [40].
2
,
0
0
T
dis
bat con bat
bat T
char
bat
P dt
P dt
q
q =
}
}

(1)
, ,
0
2
0
T
lz con fc con fc
H T
lz
P dt
P dt
q q
q

=
}
}

(2)
0
, 2 2
0 0 0
( )
T
grid
HS T T T
wt pv low H H bat
P dt
P P dt E M dt P dt
q =
+ + +
}
} } }

(3)
0
T
inj inj
inv inv
E P dt =
}

(4)
The second one, which consists in a short simulation of
twenty seconds, was performed in order to test the dynamic
performance of the energy sources and the ANFIS-based
inverter control system. In this case, this controller was also
compared with the inverter control system based on PI
controllers. The indexes used in this comparison were the
following ones [16]: mean relative error (MRE), integral time
absolute error (ITAE), and integral time square error (ITSE).
1
ref inj T
inv inv
inj
t inv
P P
P
MRE
T
=

(5)
( )
0
T
ITAE t e t dt =
}

(6)
( )
2
0
T
ITSE te t dt =
}

(7)
Figs. 8-10 show the simulation results obtained in the first
simulation, whereas Figs. 11-13 correspond to the second one.
Fig. 8 shows the WT and PV power during a week. The PV
and WT DC-DC converters incorporate a MPPT controller, so
that the renewable energy sources are generating the
maximum available power in every period. The wind speed
and the sun irradiance used in the simulations were collected
from a weather station located in Algeciras, Cdiz. The
controls of PV and WT are considered independent of the
supervisory control systems. Furthermore, Fig. 8 shows the
maximum available power in the battery, hydrogen and the
whole hybrid system, and the power demanded by the grid
during the same period. The maximum available power is
calculated as the sum of the power generated by the renewable
sources (WT and PV) and the maximum available power in
the hydrogen and battery. Note that the power demanded by
the grid is always less or equal than the maximum available
power.

Fig. 8. a) WT power and PV power; b) maximum available power in the
hydrogen and battery; c) maximum available power and power demanded by
the grid during a week.
Figure 9 depicts the duration curves (power versus
probability) of the battery and hydrogen powers during one
year, and the battery and hydrogen powers during a certain
week, the same that shown in Fig. 8. The stored power is
considered negative, so that the negative values of the
hydrogen power correspond to the electrolyzer and the
positive ones to the FC. Fig. 8c shows a zoom between the
hours 7595 and 7605. It can be observed during this period
that, when the demanded power is almost equal to the
maximum available power, the battery and FC help to the
primary energy sources by generating extra power, so that the
battery SOC and hydrogen level decrease, as shown in Fig.
10c. On the contrary, when the demanded power is very small
compared with the maximum available power, the power
excess is stored in the battery or used in the electrolyzer to
produce hydrogen, and then, the battery SOC and hydrogen
tank level increase.
The right control of the battery SOC and hydrogen level
achieved by both supervisory control systems can be checked
in Fig. 10. It shows the battery SOC and hydrogen level during
a year, and the detail of one week. In the case of the battery,
the supervisory control systems avoid a SOC under 20%, as
recommended by the battery manufacturer [41]. The classical
control system achieves higher battery SOC than the ANFIS
control system. The hydrogen tank is completely discharged
with both control systems, since it is allowed to discharge
completely. The complete discharge of the hydrogen tank is
much more frequent with the classical control system.
Otherwise, in the ANFIS control, it can be checked the
relationship between the periods of the year with high solar
radiation and the levels of the battery SOC and tank hydrogen.
7520 7540 7560 7580 7600 7620 7640 7660 7680
0
500
1000
1500
2000
a)
Time (h)
P
o
w
e
r

(
W
)


WT power
PV power
7520 7540 7560 7580 7600 7620 7640 7660 7680
0
100
200
300
400
500
b)
Time (h)
P
o
w
e
r

(
W
)


Maximum available power in the battery (ANFIS)
Maximum available power in the battery (STATES)
Maximum available power in the hydrogen (ANFIS)
Maximum available power in the hydrogen (STATES)
7520 7540 7560 7580 7600 7620 7640 7660 7680
0
1000
2000
3000
4000
5000
6000
c)
Time (h)
P
o
w
e
r

(
W
)


Maximum available power(ANFIS)
Maximum available power (STATES)
Demanded power (ANFIS)
Demanded power (STATES)
7595 7600 7605
0
2000


Copyright (c) 2013 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

8
In fact, since around the hour 2100 (early April) until the hour
6500 (late September), these levels are higher than during the
rest of the year. In the classical control, the battery SOC is
higher during the periods of the year with high solar radiation,
whereas the hydrogen level is very similar throughout the
year.
Fig. 9. a) Duration curve of the battery power during one year; b) duration
curve of the hydrogen (FC and electrolyzer) power during one year; and c)
detail of powers during one week.

Fig. 10. a) Battery SOC during one year simulation; b) hydrogen level during
one year; and c) detail of battery SOC and hydrogen level during one week.
Table I presents the results (battery, hydrogen system, and
hybrid system efficiencies and the energy injected into grid)
obtained by each control during the first simulation. Analyzing
the results, it can be concluded that the ANFIS control
achieves higher battery and hybrid system efficiencies, and it
is capable of injecting more energy into grid than the classical
control.
Figs. 11-13 show in detail the dynamic behavior of the
supervisory controls and inverter control systems.

Parameter ANFIS STATES
Battery efficiency, bat (%) 68.1 62.5
Hydrogen system efficiency, H2 (%) 37.3 38.2
Hybrid system efficiency, HS (%) 51.1 50.7
Energy injected into grid, Einv
inj
(kWh) 4800.9 4693.5
Table I. Summary of results obtained in the first simulation.

Fig. 11. a) Power demanded by the grid, PV power and WT power; b) FC,
electrolyzer and battery power with the ANFIS-based supervisory system; and
c) FC, electrolyzer and battery power with the supervisory control system
based on states.
Fig. 11 presents the power demanded by the grid and the
power of each energy source during twenty seconds. As
mentioned previously, the renewable sources are generating
the maximum available power. During the first ten seconds,
the renewable power (PV and WT power) is higher than the
power demanded by the grid. In this case, the power excess is
stored in the battery and used in the electrolyzer to produce
hydrogen. From the second ten the power demanded by the
grid increases, and it becomes higher than the renewable
power. Then, the FC and battery generate extra power to
satisfy the demanded power. Because the battery presents
quick dynamic response, during the demanded power change
0 10 20 30 40 50 60 70 80 90 100
-1000
-500
0
500
Probability (%)
P
o
w
e
r

(
W
)
a)


P
bat,ANFIS
P
bat,STATES
0 10 20 30 40 50 60 70 80 90 100
-500
-200
0
200
500
Probability (%)
P
o
w
e
r

(
W
)
b)


P
H2,ANFIS
P
H2,STATES
7520 7540 7560 7580 7600 7620 7640 7660 7680
-1000
-500
0
500
Time (h)
P
o
w
e
r

(
W
)
c)


P
bat,ANFIS
P
bat,STATES
P
H2,ANFIS
P
H2,STATES
1000 2000 3000 4000 5000 6000 7000 8000
0
20
40
60
80
100
Time (h)
%
a)


SOC
ANFIS-S
SOC
STATES
1000 2000 3000 4000 5000 6000 7000 8000
0
20
40
60
80
100
Time (h)
%
b)


H2
ANFIS-S
H2
STATES
7520 7540 7560 7580 7600 7620 7640 7660 7680
0
20
40
60
80
100
120
Time (h)
%
c)


SOC
ANFIS-S
SOC
STATES
H2
ANFIS-S
H2
STATES 7595 7600 7605
0
50


0 2 4 6 8 10 12 14 16 18 20
0
500
1000
a)
Time (s)
P
o
w
e
r

(
W
)


Demanded power PV power WT power
0 2 4 6 8 10 12 14 16 18 20
-400
-200
0
200
400
600
800
b)
Time (s)
P
o
w
e
r

(
W
)


Battery power FC power LZ power
0 2 4 6 8 10 12 14 16 18 20
-400
-200
0
200
400
600
800
c)
Time (s)
P
o
w
e
r

(
W
)


Battery power FC power LZ power
9.8 10 10.2 10.4 10.6
0
200
400
600


10 10.2 10.4 10.6
0
200
400
600


Copyright (c) 2013 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

9
(second ten), it has to deliver an additional power in order to
be able to control the DC bus voltage, while providing, in
collaboration with the FC, the power demanded by the grid.

Fig. 12. a) Power references and powers injected by the inverter into the grid;
and b) direct/quadrature-axis current components of the inverter.

Fig. 13. a) Inverter output current with ANFIS controller; b) inverter output
current with PI controllers; and c) DC bus voltage.
Fig. 12 depicts the active and reactive power references and
the active and reactive powers injected by the inverter into the
grid. In this case, although the results achieved by both
controls (ANFIS-based inverter control and PI-based inverter
control) are quite similar, the dynamic response of the inverter
controlled with the ANFIS controller presents a better
dynamic response than with the PI controller. Furthermore,
this figure shows the relationship between the direct axis
current component and the active power, and that between the
quadrature axis current component and the reactive power.
The three-phase waveform of the inverter current and the
DC bus voltage are shown in Fig. 13. It can be seen that the
control implemented in the battery converter makes possible
to maintain constant the DC bus voltage during all the
simulation. Otherwise, with the ANFIS controller, the inverter
output current presents a total harmonic distortion (THD) of
7.15% during the first ten seconds, and 4.1% during the last
ten seconds.
Table II depicts the values of the indexes (MRE, ITAE,
ITSE) and THD of the inverter output current obtained with
each controller during the second simulation. According to the
results, the ANFIS controller shows a better inverter control,
since the error indexes and THD are lower than those obtained
by the PI controllers.

Parameter ANFIS controller PI controller
MRE (Active power control) 7.8e-4 9.6e-4
ITAE (Active power control) 0.207 0.228
ITSE (Active power control) 0.024 0.026
MRE (Reactive power control) -9e-4 -12e-4
ITAE (Reactive power control) 0.096 0.108
ITSE (Reactive power control) 59e-4 71e-4
THD (Current injected, 0-10s) 7.15% 7.71%
THD (Current injected, 10-20s) 4.11% 4.29%
Table II. Summary of results obtained in the second simulation seconds.
VI. CONCLUSION
This paper has presented and evaluated an ANFIS-based
EMS of a grid-connected hybrid system, which is composed
of renewable energy sources (WT and PV panels), hydrogen
(FC, electrolyzer and hydrogen tank) and battery.
The renewable energy sources operate at the MPP,
independently of the EMS. Therefore, the EMS is responsible
for managing the operation of the ESS (battery and hydrogen
system), taking into account the power demanded by the grid,
the available renewable power, and the available ESS power
(battery SOC and hydrogen tank level). Once determined
which ESS has to absorb/generate the power excess/deficit,
the appropriate control of the DC/DC converter associated to
the battery and hydrogen system allows the ESS to achieve the
operation required by the EMS, maintaining constant the DC
bus voltage.
On the other hand, in coordination with the EMS, the three-
phase inverter is controlled by an ANFIS-based controller in
order to regulate the active and reactive power that the hybrid
system is required to deliver with the grid.
This EMS has been evaluated and compared with a classical
EMS composed of state-based supervisory control system and
inverter control system based on PI controllers, by large scale
operation (one year simulation) and by short scale operation
(twenty seconds simulations). During the large-scale operation
it can be observed that both EMS are able to supply the energy
required by the grid, maintaining the battery SOC within the
0 2 4 6 8 10 12 14 16 18 20
0
200
400
600
800
1000
a)
Time (s)
P
o
w
e
r

(
W
)


P
ref
inv
P
inj
inv,ANFIS-I
P
inj
inv,PI
0 2 4 6 8 10 12 14 16 18 20
0
0.5
1
1.5
2
2.5
3
b)
Time (s)
C
u
r
r
e
n
t

(
A
)


d-axis current ANFIS-I
d-axis current PI
q-axis current ANFIS-I
q-axis current PI
9.95 10 10.05 10.1 10.15
400
600
800


Q
ref
inv
Q
inj
inv,ANFIS-I
Q
inj
inv,PI
9.95 10 10.05 10.1 10.15
0
100
200
300


9.98 10 10.02 10.04 10.06 10.08
0
1
2
3


9.9 9.95 10 10.05 10.1 10.15
-5
0
5
a)
Time (s)
C
u
r
r
e
n
t

(
A
)


Inverter output current ANFIS-I
9.9 9.95 10 10.05 10.1 10.15
-5
0
5
b)
Time (s)
C
u
r
r
e
n
t

(
A
)


Inverter output current PI
0 2 4 6 8 10 12 14 16 18 20
22
24
26
28
30
c)
Time (s)
V
o
l
t
a
g
e

(
V
)


DC bus voltage ANFIS-I
DC bus voltage PI
10 10.5 11
23.5
24
24.5
25


Copyright (c) 2013 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

10
operational limits. Nevertheless, the ANFIS-based EMS
achieves better results, since it presents higher battery and
hybrid system efficiencies, and it is capable of injecting more
energy into grid than the classical EMS. The short scale
operation shows the right dynamic response of the hybrid
system and inverter against sudden power variations and its
effect on the DC bus voltage and three-phase waveform of the
inverter output voltage and current. In this case, the inverter
controlled with ANFIS presents lower error indexes (MRE,
ITAE, ISTE) and THD than those obtained with the PI
controllers.
The results of both simulations demonstrated that the
proposed EMS allows a better control than the classical EMS
and reliable operation in grid-connected applications.
REFERENCES
[1] M. P. Kazmierkowski, M. Jasinski, and G. Wrona, DSP-based
control of grid-connected power converters operating under grid
distortions, IEEE Trans. Ind. Informat., vol. 7, no. 2, pp. 204-2011,
May 2011.
[2] P. Siano, C. Cecati, H. Yu, and J. Kolbusz, Real time operation of
smart grids via FCN networks and optimal power flow, IEEE Trans.
Ind. Informat., vol. 8, no. 4, pp. 944-952, Nov. 2012.
[3] Z. Nie, X. Xiao, R. McMahon, P. Clifton, Y. Wu, and S. Shao,
Emulation and control methods for direct drive linear wave energy
converters, IEEE Trans. Ind. Informat., vol. 9, no. 2, pp. 790-798,
May. 2013.
[4] F. Kennel, D. Grges, and S. Liu, Energy management for smart
grids with electric vehicles based on hierarchical MPC, IEEE Trans.
Ind. Informat., vol. PP, no. 99, pp. 1-9, 2012.
[5] M. N. Marwali, J. W. Jung, and A. Keyhani, Control of distributed
generation systems - Part II: Load sharing control, IEEE Trans.
Power Electr., vol. 19, no. 6, pp. 1551-1561, Nov. 2004.
[6] P. Bajpai, and V. Dash, Hybrid renewable energy systems for power
generation in stand-alone applications: A review, 2012, Renew. Sust.
Energ. Rev., vol. 16, no. 5, pp. 2926-2939, Jun. 2012.
[7] S. Jiang, W. Wang, H. Jin, and D. Xu, Power management strategy
for microgrid with energy storage system, in 37th Annual
Conference on IEEE Industrial Electronics Society (IECON 2011),
pp. 1524-1529.
[8] L. Valverde, F. Rosa, and C. Bordons, Design, planning and
management of a hydrogen-based microgrid, IEEE Trans. Ind.
Informat., vol. 9, no. 3, pp. 1398-1404, Aug. 2013.
[9] S. N. Bhaskara, and B. H. Chowdhury, Microgrids A review of
modeling, control, protection, simulation and future potential, in
IEEE Power and Energy Society General Meeting 2012, pp. 1-7.
[10] M. Jamil, B. Hussain, M. Abu-Sara, R. J. Boltryk, and S. M. Sharkh,
Microgrids power electronic converters: State of the art and future
changes, in Proceedings of the 44th International Universities
Power Engineering Conference (UPEC 2009), pp. 1-5.
[11] F. Gonzlez-Espn, I. Patrao, E. Figueres, and G. Garcer. An
adaptive digital control technique for improved performance of grid
connected inverters, IEEE Trans. Ind. Informat., vol. 9, no. 2, pp.
708-718, May. 2013.
[12] M. P. Kazmierkowski, M. Jasinski, and G. Wrona. DSP-based
control of grid-connected power converters operating under grid
distortions, IEEE Trans. Ind. Informat., vol. 7, no. 2, pp. 204-211,
May 2011.
[13] S. Dasgupta, S. N. Mohan, S. K. Sahoo, and S. K. Panda, A plug
and play operational approach for implementation of an autonomous-
micro-grid system, IEEE Trans. Ind. Informat., vol. 8, no. 3, pp.
615-629, Aug. 2012.
[14] J. L. Bernal-Agustn, and R. Dufo-Lpez, Simulation and
optimization of stand-alone hybrid renewable energy systems,
Renew. Sust. Energ. Rev., vol. 13, no. 8, pp. 2111-2118, Oct. 2009.
[15] M. A. Akcayol, Application of adaptive neuro-fuzzy controller for
SRM, Adv. Eng. Softw., vol. 35, no. 3-4, pp. 129-137, Mar. 2004.
[16] N. Altin, and I. Sefa dSPACE based adaptive neuro-fuzzy controller
of grid interactive inverter, Energ. Convers. Manage., vol. 56, pp.
130-139, May. 2012.
[17] M. Singh, and A. Chandra, Real-time implementation of ANFIS
control for renewable interfacing inverter in 3P4W distribution
network, IEEE Trans. Ind. Electron., vol. 60, no. 1, pp. 121-128,
Jan. 2013.
[18] SimPowerSystems, Reference, Hydro-Qubec, MathWorks, Inc.,
Natick, MA, 2012.
[19] J. I. Rosell, and M. Ibez, Modelling power output in photovoltaic
modules for outdoor operating conditions, Energ. Convers.
Manage., vol. 47, no. 15-16, pp. 2424-2430, Sep. 2006.
[20] M. G. Villalva, J. R. Gazoli, and E. R. Filho, Comprehensive
approach to modeling and simulation of photovoltaic arrays, IEEE
Trans. Ind. Electron., vol. 24, no. 5, pp. 1198-1208, May. 2009.
[21] M. A. S. Masoum, H. Dehbonei, and E. F. Fuchs, Theoretical and
experimental analyses of photovoltaic systems with voltage and
current-based maximum power-point tracking, IEEE Trans. Energy
Conver., vol. 17, no. 4, pp. 514-522, Dec. 2002.
[22] D. Stolten, Hydrogen and Fuel Cells: Fundamentals, Technologies
and Applications. Weinheim: Wiley-VCH, 2010.
[23] M. Uzunoglu, O. C. Onar, and M. S. Alam. Modeling, control and
simulation of a PV/FC/UC based hybrid power generation system for
stand-alone applications, Renew. Energ., vol. 34, no. 3, pp. 509-520,
Mar. 2009.
[24] B. Panahandeh, J. Bard, A. Outzourhit, and D. Zejli. Simulation of
PV-wind hybrid systems combined with hydrogen storage for rural
electrification, Int. J. Hydrogen Energ., vol. 36, no. 6, pp. 4185-
4197, Mar. 2011.
[25] J. T. Pukrushpan, A. G. Stefanopoulou, and H. Peng, Control of fuel
cell power systems: Principles, modeling, analysis, and feedback
design. London: Springer Verlag, 2002.
[26] P. Garca, L. M. Fernndez, C. A. Garca, and F. Jurado,
Comparative study of PEM fuel cell models for integration in
propulsion systems of urban public transport, Fuel Cells, vol. 10, no.
6, pp. 1024-1039, Dec. 2010.
[27] L. M. Fernndez, P. Garca, C. A. Garca, J. P. Torreglosa, and F.
Jurado, Comparison of control schemes for a fuel cell hybrid
tramway integrating two dc/dc converters, Int. J. Hydrogen Energ.,
vol. 35, no. 11, pp. 5731-5744, Jun. 2010.
[28] P. Garca, J. P. Torreglosa, L. M. Fernndez, and F. Jurado,
Viability study of a FC-battery-SC tramway controlled by
equivalent consumption minimization strategy, Int. J. Hydrogen
Energ., vol. 37, no. 11, pp. 9368-9382, Jun. 2012.
[29] J. P. Torreglosa, P. Garca, L. M. Fernndez, and F. Jurado.
Predictive control for the energy management of a fuel cell-battery-
supercapacitor tramway, IEEE Trans. Ind. Informat., Vol. PP, no.
99, pp. 1-9, 2013.
[30] P. Garca, J. P. Torreglosa, L. M. Fernndez, and F. Jurado, Control
strategies for high-power electric vehicles powered by hydrogen fuel
cell, battery and supercapacitor, Expert Syst. Appl., vol. 40, no. 12,
pp. 4791-4804, Sep. 2013.
[31] A. Tani, M. B. Camara, B. Dakyo, and Y. Azzouz, DC/DC and
DC/AC converters control for hybrid electric vehicles energy
management-ultracapacitors and fuel cell, IEEE Trans. Ind.
Informat., vol. 9, no. 3, pp. 686-696, May. 2013.
[32] D. Linden, and T. B. Reddy, Handbook of Batteries. New York:
McGraw-Hill, 2002.
[33] O. Tremblay, and L. A. Dessaint, Experimental validation of a
battery dynamic model, World Electric Vehicular Journal, vol. 3,
no. 1, pp. 1-10, 2009.
[34] M. K. Kazimierczuk, Pulse-Width Modulated DCDC Power
Converters. New York: John Wiley and Sons, 2008.
[35] A. T. Azar, Fuzzy Systems. Viena: In Tech, 2010.
[36] J.-S. R. Jang, ANFIS: adaptive-network-based fuzzy inference
system, IEEE Trans. Syst. Man Cyb., vol. 23, no. 3, pp. 665 - 685,
May/Jun. 1993.
[37] J.-S. R. Jang, and C. T. Sun, Neuro-fuzzy modeling and control,
Proceedings of the IEEE, vol. 83, no. 3, pp. 378-406, Mar.1995.
[38] J.-S. Jang, C. T. Sun, and E. Mizutani, Neuro-Fuzzy and soft
computing. Prentice-Hall: Englewood Cliffs, NJ, 1997.
[39] Fuzzy Logic Toolbox, Reference, Hydro-Qubec, MathWorks, Inc.,
Natick, MA, 2012.
Copyright (c) 2013 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TII.2013.2290069, IEEE Transactions on Industrial Informatics
TII-13-0441

11
[40] M. Castaneda, A. Cano, F. Jurado, H. Sanchez, and L M. Fernandez,
Sizing optimization, dynamic modeling and energy management
strategies of a stand-alone PV/hydrogen/ battery-based hybrid
system, Int. J. Hydrogen Energ., vol. 38, no. 10, pp. 3830-3845,
Apr. 2013.
[41] BAE, Energy from Batteries, BAE Secura PVS solar, 2012.
[Online]
http://www.bornay.com/userfiles/descargas/BAE_PVS_Cell_engl_20
12_12.pdf.







Pablo Garcia was born in La Linea de la
Concepcion (Cdiz), Spain, in 1984. He received the
MSc degree in Engineering from the University of
Cadiz, Spain, in 2007, and the Ph.D. degree from the
University of Cadiz, Spain, in 2010.
Since 2008, he is Professor at the Department of
Electrical Engineering of the University of Cadiz,
Spain. His research interest focuses on power
systems and power management in hybrid systems.










Carlos Andrs Garca was born in La Linea de la
Concepcin (Cadiz), Spain. He received the M.Sc.
degree in engineering and Ph.D. degree from the
University of Cadiz, Cadiz, Spain, in 2004 and 2009,
respectively.
Since 1988, he has been a Professor with the
Department of Electrical Engineering, University of
Cadiz. His research interest focuses in electric
machines and renewable energy.





Luis M. Fernndez (M11) was born in Los Barrios
(Cadiz), Spain. He received the M.Sc. degree in
electrical engineering from the University of Seville,
Seville, Spain, in 1997 and the Ph.D. degree from
the University of Cadiz, Cadiz, Spain, in 2004.
From 1997 to 2000, he was with the Department of
Development and Research, Desarrollos Eolicos
S.A. In 2000, he joined the University of Cadiz,
Spain, where he is currently an Associate Professor
in the Department of Electrical Engineering. His
current research interests include renewable energy
sources and fuel cell systems.




Francisco Llorens was born in Algeciras (Cadiz),
Spain in 1967. He received the M.Sc. degree in
electrical engineering and Ph.D. degree from the
University of Seville, Seville, Spain, in 1994 and
2010, respectively.
Since 1994, he has been with the Department of
Electrical Engineering, University of Cdiz, Cdiz,
Spain, where he is currently an Associate Professor.
His primary areas of interest are distribution power
optimization and renewable energy.




Francisco Jurado (M00-SM06) was born in Linares
(Jaen), Spain. He received the M.Sc. and Dr. Ing.
degrees from the National University of Distance
Education, Madrid, Spain, in 1995 and 1999,
respectively.
Since 1985, he is a Professor with the Department of
Electrical Engineering, University of Jaen, Linares,
Spain. His research activities have been devoted to
several topics, e.g., power systems, modeling, and
renewable energy.