Control Engineering Practice 81 (2018) 215–230

Contents lists available at ScienceDirect

Control Engineering Practice

journal homepage: www.elsevier.com/locate/conengprac

Intelligent droop control and power management of active generator for

ancillary services under grid instability using fuzzy logic technology
Youssef Krim a ,∗, Dhaker Abbes b , Saber Krim a , Mohamed Faouzi Mimouni a
a Research Unit of Industrial Systems Study and Renewable Energy, Electrical Engineering Department, National Engineering School of Monastir, Avenue Ibn El Jazzar,
5019, University of Monastir, Tunisia
b
Laboratory of Electrical Engineering and Power Electronics of Lille (L2EP), Ecole des Hautes Etudes d’Ingénieur (HEI), 13, rue de Toul, F-59046 Lille, France

ARTICLE INFO ABSTRACT

Keywords: In this paper, a control and power supervisor for a flexible operation of a Renewable Distributed Generator
Adaptive fuzzy logic droop control (RDG) is introduced. This RDG consists of a combination of a wind system and a hybrid storage system made up
Fuzzy logic supervisor of Batteries (BT) and Super-Capacitors (SC). RDG is associated with a load and a fluctuating grid to form an Active
Active generator
Generator (AG). According to the grid fluctuation, AG can operate in a grid-connected and standalone mode. The
Hybrid energy storage system
objective of this work is to investigate a novel control strategy for AG integrated into the grid in order to maintain
Fuzzy logic islanding detection
its voltage and frequency in an allowable range and to ensure the continuity of the power supply in case of a
grid fault. The structure of the proposed control strategy consists of a Fuzzy Logic Supervisor (FLS), an adaptive
Fuzzy Logic Droop Control (FLDC) and a Fuzzy Logic Islanding Detection (FLID). FLS is developed to manage
the power flows between the storage devices by choosing the optimal operating mode, thereby ensuring the grid
stability and the continuous supply of the load by maintaining the state of charge of SC and BT at acceptable
levels and to reduce stresses on BT and improve their life cycle. FLID is used to detect de standalone mode in case
of grid failure. Finally, FLDC is used to control the active and reactive powers exchanged with the grid, ensuring
its stability by maintaining its frequency and its voltage in optimal margins. The effectiveness of the proposed
control method is validated by simulation results and compared with a generalized control technique.

1. Introduction
The significant increase in the population and areas of using electri-
cal energy expected in the last few years must cope with the growing
energy consumption associated with it. To do this, this evolution
must result in intelligent management of an electrification network
(Petronela et al., 2016). Therefore, the power grid is becoming more
and more interconnected and meshed. One of the envisaged solutions
concerns the reinforcement of the electricity supply network by the
integration of innovative energy storage devices, new renewable and
relocated production means, as well as an energy optimization of the
grid architecture (Glasnovic & Margeta, 2011). The wind technology
has become a favored form of renewable energy technologies because
it is seen as clean and sustainable (Islam, Djamel, Abdel-Moumen, Bilal,
& Ikram. El, 2018). The availability of wind energy depends on the
climatic and geographical contexts of the installation region, which
represents the major problem of wind turbines. Therefore, it is naturally
very unlikely to have a concomitance between production and demand.
In order to guarantee production, the aggregation of substation storage
units will make it possible to overcome this problem. Hybrid Energy
Storage Systems (HESSs) have become more and more important in
Renewable Distributed Generators (RDGs) (Jia, Mu, & Qi, 2014). The
wind technology application stresses the Batteries (BT) storage system,
because it uses a large fraction of the energy stored in it (Matthieu,
Arnaud, Karl, Moe, Marc, & Richard, 2017). The BT lifespan is limited
by the charge/discharge cycle. Compared to BT, the Super-Capacitors
(SC) storage system life is much longer and has much higher power
density. Moreover, SC can provide a fast and effective energy output
because of its high power density and high efficiency (Wang, Liu,
Pan, & Chen, 2017). Thus, the adoption of SC in HESS is an effective
solution to prolong the lifespan of BT in renewable energy production
applications. The HESS technology can be passive, semi-active or purely
active. Passive HESS is the simplest configuration, such as SC and
BT directly connected to a DC bus without any power converter, and
is characterized by its low cost. However, the performance of this
configuration is limited because SC cannot be used effectively (Ziyou,
Jun, Heath, Jianqiu, & Minggao, 2017). Semi-active HESS uses a single
power converter. This topology combines good performance and low
system cost (Ziyou et al., 2017). The purely active HESS configuration
uses two DC/DC converters. It gives the possibility to control SC and
BT currents simultaneously. Therefore, this configuration combines

∗ Corresponding author.
E-mail address: krim_enim@hotmail.com (Y. Krim).

https://doi.org/10.1016/j.conengprac.2018.09.013
Received 21 January 2018; Received in revised form 8 July 2018; Accepted 12 September 2018
Available online xxxx
0967-0661/© 2018 Elsevier Ltd. All rights reserved.
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

efficiency and good performance. For this reason, the purely active 3. Renewable distributed generator model and control
configuration is focused on in this study.
The key problem with HESS is its protection against over-discharge 3.1. Model of wind generator
and overcharge, as well as the BT protection against the rapid fluctu-
ation of charge/discharge cycles using SC. For this reason, the major The aerodynamic power produced by the turbine can be expressed
task is to optimize the power flow management between BT and SC, by (Youssef, Saber, & Mohamed, 2017):
which manages the reference current for each storage device. The most
common of Energy Management Strategies (EMSs) are rule-based EMSs 𝑃𝑎𝑒𝑟 = 0.5𝜌𝜋𝑅2 𝑣3 𝐶𝑝 (𝜆, 𝛽) (1)
(Song et al., 2014). However, focusing on renewable energy, if RDG has
where R is the radius of the rotation of each blade, 𝜌 ≈ 1, 2 kg/m3
to operate on different modes, rule development will become complex
denotes the air density, v is the wind speed, and 𝐶𝑝 can then be
due to several variables and case studies. In this framework, fuzzy logic
represented as a polynomial function of 𝜆 and 𝛽. In our case, this
appears as a promising tool owing to the possibility of avoiding most
polynomial is defined by (Youssef et al., 2017):
of the mathematical stiffness and complexity in problem formulation ( ) ( )
and representing it based on human reasoning (Victor et al., 2016). 151 18.4
𝐶𝑝 (𝜆, 𝛽) = 0.53 − 0.58𝛽 − 0.002𝛽 2.14 − 10 exp − (2)
EMS based on fuzzy logic is used in several applications, such as the 𝜆𝑖 𝜆𝑖
energy management of HESS used in a tramway (Victor et al., 2016) and 1
𝜆𝑖 = (3)
EMS for standalone RDGs (Erdinc & Uzunoglu, 2011). In this study, EMS 1
− 0.003
𝜆−0.02𝛽 𝛽 3 +1
based on a Fuzzy Logic Supervisor (FLS) is proposed to monitor RDG in
standalone and grid-connected modes. The purpose of suggested FLS is We deduce the expression of the generated aerodynamic torque as
to ensure a balance between production and consumption, to guarantee follows:
a continuous load supply by maintaining the State Of Charge (SOC) 𝑃𝑎𝑒𝑟 0.5𝜌𝜋𝑅2 𝑣3 𝐶𝑝 (𝜆, 𝛽)
of SC (𝑈𝑠𝑐 ) and BT (SOCbat ) at acceptable levels, and to assist RDG to 𝑇𝑎𝑒𝑟 = = (4)
𝛺𝑚 𝛺𝑚
contribute to the improvement of electrical network performance.
On the other hand, the studied RDG is associated with a load to This torque makes it possible to turn the rotor of PMSG, which is
form an Active Generator (AG). AG can operate according to the grid modeled in the park frame by the following expression (Masmoudi,
stability in two main operation modes. First, the grid connection mode Abdelkafi, & Krichen, 2011):
participates in system services by adjusting the frequency and amplitude 𝑑𝑖𝑠𝑑
of the grid voltage. Second, the standalone mode ensures a continuous 𝑉𝑠𝑑 = 𝐿𝑠 + 𝑅𝑠 𝑖𝑠𝑑 − 𝑝𝛺𝑚 𝑖𝑠𝑞
𝑑𝑡
power supply of the load in case of a grid fault. There is an intermediate 𝑑𝑖𝑠𝑞 (5)
mode to safety reconnecting AG to the utility grid. Thus, some research 𝑉𝑠𝑞 = 𝐿𝑠 + 𝑅𝑠 𝑖𝑠𝑞 + 𝑝𝛺𝑚 𝑖𝑠𝑑 + 𝑝𝛺𝑚 𝜙𝑚
𝑑𝑡
work has processed the regulation of the frequency and voltage in a 𝑇𝑒𝑚 = 𝑝𝜙𝑚 𝑖𝑠𝑑
connected or standalone mode. Table 1 show that the droop control
method is commonly employed in RDG, which has more than one Fig. 2 illustrates a vector control of PMSG in the PARK frame. This
operation mode. strategy consists in keeping axis d constantly aligned with the flux vector
Consequently, this study is focused on a novel control technique of the magnet. The reference of direct current 𝑖𝑑 is kept at zero. The
based on EMS using FLS and adaptive FLDC, capable of monitoring AG reference for quadratic current 𝑖𝑞 is determined by the electromagnetic
in different operation modes and injecting active and reactive powers torque deduced by the Maximum Power Point Tracking (MPPT) strategy
into the grid with high precision ensuring its stability. as follows (Krim, Abbes, Krim, & Mimouni, 2017):
This paper is structured as follows. Section 2 provides an overview 5 3
𝑇𝑒𝑚−𝑀𝑃 𝑃 𝑇 𝑃 1 𝜌𝐶𝑝 max 𝑅 𝛺 𝑚
of the wind hybrid generator configuration. Section 3 presents the 𝑖∗𝑠𝑞 = 𝑤𝑖𝑡ℎ 𝑇𝑒𝑚−𝑀𝑃 𝑃 𝑇 = 𝑀𝑃 𝑃 𝑇 = (6)
𝑝𝜙𝑚 𝛺𝑚 2 3
modeling and control of RDG. Section 4 discusses the working principle 𝜆𝑜𝑝𝑡
of EMS based on FLS. Section 5 focuses on the FLDC technique. The
simulation results and the discussion are given in Section 6. Section 7 3.2. BT model and control
concludes this report.
We choose for a main storage unit a BT bank, which must be
2. Active generator configuration connected to the DC bus through a bidirectional converter (DC/DC
converter 3). We have retained the recent technology of the CIEMAT
Before undertaking modeling, we need to define in more detail the (Research Center for Energy, Environment and Technology, Espagne)
architectures of the different parts that make up RDG, detailed in Fig. 1: BT model, which has very high energy density. An equivalent model for
a wind generator, SC, BT, and a DC bus link. RDG is associated with BT is depicted in Fig. 3. It is composed by voltage source 𝐸𝑏 in series
a balanced AC load to form AG. The latter can operate in a grid- with internal resistance 𝑅𝑖 .
connected mode or an islanded one, according to the stability status The expressions of BT quantities are expressed in the following
of the main power grid. In addition, we need to choose the most (Cabrane, Ouassaid, & Maaroufi, 2017):
appropriate converter type to control each AG component, which will − The general expression of the BT voltage is as follows:
ensure the adaptation of this one to the continuous DC bus. The three-
bladed wind generator absorbs mechanical power 𝑃𝑚 , captured by wind 𝑉𝑏𝑎𝑡 = 𝑛𝑏 𝐸𝑏 + 𝑛𝑏 𝑅𝑖 𝑖𝑏𝑎𝑡 (7)
speed v passing through surface S covered by its blades. The turbine
− The expression of SOC is as follows:
then generates torque 𝑇𝑚 , which will drive at angular velocity 𝜔𝑚 the
rotor of a rotating machine. This transforms the absorbed mechanical 𝑄𝑑
𝑆𝑂𝐶 = 1 − (8)
power into exploitable electrical power. We choose to use a Permanent 𝐶𝑏𝑎𝑡
Magnets Synchronous Generator (PMSG), able to operate at different Capacity 𝐶𝑏𝑎𝑡 of BT is expressed as a function of charging and
speeds until it stops without stalling, which require little maintenance. discharging current 𝑖𝑏𝑎𝑡 .
To complete the power lack or excess, BT/SC HESS having two types
𝐶𝑏𝑎𝑡 1.67
of storage units is suggested. The role of this AG is double: maintaining = ( )0.9 (1 + 0.005𝛥𝑇 ) (9)
the frequency and voltage of the main grid in the desired stability range 𝐶10 𝑖
1 + 0.67 𝑖𝑏𝑎𝑡
and ensuring the continuous supply of the load in case of a grid failure. 10

Indeed, a new control strategy based on FLS and FLDC will be applied where charge/discharge current 𝑖10 corresponds to rated capacity 𝐶10 .
to make the studied AG intelligent and able to meet the aforementioned Fig. 5(a) represents the control of DC/DC converter 3 and the BT
objectives. current:

216
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Fig. 1. Configuration of proposed hybrid system by using combination of BT and SC.

Fig. 2. PMSG vector control.

217
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Table 1
Summary of different droop control methods. (See Furtado et al. (2008), Guerrero et al. (2006), Hou et al. (2015), Il-Yop et al. (2010), Juan et al.
(2009), Lee et al. (2009), Matas et al. (2010), Sahyoun et al. (2015), Shuai et al. (2016), Tayaba et al. (2017), Tolani and Sensarma (2012).).

− The converter control is made by the duty cycle defined as follows:

𝑈𝑚𝑏𝑎𝑡
𝑚𝑏𝑎𝑡 = (11)
𝑈𝐷𝐶

3.3. SC model and control

Faced with the slow dynamics and the risk of the premature wear of
BT, we add a secondary storage unit to the system, intended to absorb
or provide power peaks. The technology of SC is chosen. The model of
Fig. 3. Equivalent model of nb BT elements in series.
SC used in this work is presented in Fig. 4(a). An SC module composed
of multiple in-series and parallel SC is thus modeled by capacitance 𝐶𝑠𝑐
in series with resistance 𝑅𝑠𝑐 (Fig. 4(b)):
− A PI controller is used to control the charge and discharge current
𝑁𝑠 𝑁𝑝
of BT and make it equivalent to its reference ‘‘𝑖𝑏𝑎𝑡−𝑟𝑒𝑓 ’’: 𝑅𝑠𝑐 = 𝑅𝑠𝑠𝑐 ; 𝐶 = 𝐶𝑠𝑠𝑐 (12)
𝑁𝑝 𝑠𝑐 𝑁𝑠

( ) where 𝐶𝑠𝑠𝑐 is the nominal capacitance and 𝑅𝑠𝑠𝑐 is the equivalent in-series
𝑈𝑚𝑏𝑎𝑡 = 𝑉𝑏𝑎𝑡 − 𝑃 𝐼 𝑖𝑏𝑎𝑡−𝑟𝑒𝑓 − 𝑖𝑏𝑎𝑡 (10) resistance.

218
Y. Krim et al.
Fig. 4. (a) SC model, (b) Equivalent model of SC module.
Fig. 5(b) illustrates the control of DC/DC converter 2 and the SC
current:
− Current control: A PI regulator is used to maintain the charging
and discharging current of SC equivalent to its reference value ‘‘𝑖𝑠𝑐−𝑟𝑒𝑓 ’’.
This PI controller is expressed as follows:
( )
𝑈𝑚𝑠𝑐 = 𝑈𝑠𝑐 − 𝑃 𝐼 𝑖𝑠𝑐−𝑟𝑒𝑓 − 𝑖𝑠𝑐 (13)
− Converter control: DC/DC converter 2 permits adapting the SC
output voltage to the adequate inverter input voltage. The control of
DC/DC converter 2 is given by the duty ratio as follows:
𝑈𝑚𝑠𝑐
𝑚𝑠𝑐 = (14)
𝑈𝐷𝐶
Finally, the DC bus voltage is controlled according to the principle
described in Fig. 5(c). The PI regulator calculates the reference of
DC bus current 𝑖∗𝐷𝐶 to maintain the DC bus voltage equivalent to its
reference 𝑈𝐷𝐶∗ . The DC bus voltage can be determined by the following
relationship:
𝑑𝑈𝐷𝐶
𝐶 = 𝑖𝑚𝑔 − 𝑖𝑚𝑠𝑐 − 𝑖𝑚𝑏𝑎𝑡 − 𝑖min 𝑣 (15)
𝑑𝑡
where 𝑖𝑚𝑔 , 𝑖𝑚𝑠𝑐 , 𝑖𝑚𝑏𝑎𝑡 , and 𝑖𝑚𝑖𝑛𝑣 represent the currents modeled by the
wind generator, SC, BT, and the load and the grid, respectively. C is the
capacity of the DC bus.
The HESS can also be used to regulate the DC bus voltage. In this
case, BT and SC power references ‘‘𝑃𝑠𝑐−𝑟𝑒𝑓 ’’ and ‘‘𝑃𝑏𝑎𝑡−𝑟𝑒𝑓 ’’ are calculated
by considering the ‘‘𝑃𝑔 ’’ fluctuation wind power and the required power
for the DC bus voltage regulation (𝑃DC ∗ ).
The reference currents of BT and SC (respectively 𝑖𝑠𝑐−𝑟𝑒𝑓 and 𝑖𝑏𝑎𝑡−𝑟𝑒𝑓 )
are delivered by FLS.
4. Fuzzy logic supervisor
4.1. FLS structure
The principle of determining the reference currents of BT and SC is
detailed in Fig. 6. A Low-Pass Filter (LPF) is used to filter the power
quantity to be stored (𝑃𝑠𝑡𝑜 ). The purpose of this LPF is to construct
the ‘‘ 𝑃𝑏𝑎𝑡−𝑠𝑡𝑜 ’’ power of BT and to divert the rapid fluctuations in ‘‘
𝑃𝑠𝑐−𝑠𝑡𝑜 ’’ power into SC. However, it reduces the peak power demand
and the charging/discharging cycle on BT. After that, the SC power is
determined by the difference between 𝑃𝑠𝑡𝑜 and 𝑃𝑏𝑎𝑡−𝑠𝑡𝑜 .
To select the exact reference powers of BT and SC with the considered
SOC of each storage system, four switches are used. They are controlled
by fuzzy logic as a function of CS1, CS2, CS3 and CS4.
Switches 1 and 3 are responsible for improving the system perfor-
mance, in terms of dynamic behavior of BT and their lifespan.
∙ Switch 1 allows selecting between 𝑃𝑠𝑡𝑜 and 𝑃𝑏𝑎𝑡−𝑠𝑡𝑜 :
If 𝑃𝑠𝑡𝑜 is positive and 𝑈𝑠𝑐 ≥𝑈𝑠𝑐𝑚𝑎𝑥 then 𝑃𝑏𝑎𝑡−𝐶𝑆1 = 𝑃𝑠𝑡𝑜 else 𝑃𝑏𝑎𝑡−𝐶𝑆1 =
𝑃𝑏𝑎𝑡−𝑠𝑡𝑜
If 𝑃𝑠𝑡𝑜 is negative and 𝑈𝑠𝑐 ≤ 𝑈 𝑠𝑐𝑚𝑖𝑛 then 𝑃𝑏𝑎𝑡−𝐶𝑆1 = 𝑃𝑠𝑡𝑜 else 𝑃𝑏𝑎𝑡−𝐶𝑆1 =
𝑃𝑏𝑎𝑡−𝑠𝑡𝑜
∙ Switch 3 permits choosing between 𝑃𝑠𝑡𝑜 and 𝑃𝑠𝑐−𝑠𝑡𝑜 :
219
Control Engineering Practice 81 (2018) 215–230
If 𝑃𝑠𝑡𝑜 is positive and 𝑆𝑂𝐶 𝑏𝑎𝑡 ≥90% then 𝑃𝑠𝑐−𝐶𝑆3 = 𝑃𝑠𝑡𝑜 else 𝑃𝑠𝑐−𝐶𝑆3 =
𝑃𝑠𝑐−𝑠𝑡𝑜
If 𝑃𝑠𝑡𝑜 is negative and 𝑆𝑂𝐶 𝑏𝑎𝑡 ≤ 30% then 𝑃𝑠𝑐−𝐶𝑆3 = 𝑃𝑠𝑡𝑜 else
𝑃𝑠𝑐−𝐶𝑆3 = 𝑃𝑠𝑐−𝑠𝑡𝑜
To elaborate the reference powers of BT and SC, the SOC of each
storage system must be taken in account. Switches 2 and 4 are used
to extract the exact reference powers of SC and BT by maintaining the
SOC of every storage system into acceptable levels, to protect them from
overcharge and discharge excess.
∙ Switch 2 enables choosing between 0 and 𝑃𝑏𝑎𝑡−𝐶𝑆1 :
If 𝑃𝑏𝑎𝑡−𝐶𝑆1 is positive and 𝑆𝑂𝐶 𝑏𝑎𝑡 ≥𝑆𝑂𝐶 𝑏𝑎𝑡−𝑚𝑎𝑥 then 𝑃𝑏𝑎𝑡−𝑟𝑒𝑓 = 0 else
𝑃𝑏𝑎𝑡−𝑟𝑒𝑓 = 𝑃𝑏𝑎𝑡−𝐶𝑆1 .
If 𝑃𝑏𝑎𝑡−𝐶𝑆1 is negative and 𝑆𝑂𝐶 𝑏𝑎𝑡 ≤ SOC 𝑏𝑎𝑡−𝑚𝑖𝑛 then 𝑃𝑏𝑎𝑡−𝑟𝑒𝑓 = 0 else
𝑃𝑏𝑎𝑡−𝑟𝑒𝑓 = 𝑃𝑏𝑎𝑡−𝐶𝑆1 .
∙ Switch 4 allows selecting between 0 and 𝑃𝑠𝑐−𝐶𝑆3 :
If 𝑃𝑠𝑐−𝐶𝑆3 is positive and 𝑈𝑠𝑐 ≥𝑈𝑠𝑐𝑚𝑎𝑥 then 𝑃𝑠𝑐−𝑟𝑒𝑓 = 0 else 𝑃𝑠𝑐−𝑟𝑒𝑓 =
𝑃𝑠𝑐−𝐶𝑆3 .
If 𝑃𝑠𝑐−𝐶𝑆3 is negative and 𝑈𝑠𝑐 ≤ 𝑈 𝑠𝑐𝑚𝑖𝑛 then 𝑃𝑠𝑐−𝑟𝑒𝑓 = 0 else 𝑃𝑠𝑐−𝑟𝑒𝑓 =
𝑃𝑠𝑐−𝐶𝑆3 .
Indeed, the low rate of the BT charge is 𝑆𝑂𝐶 𝑏𝑎𝑡−𝑚𝑖𝑛 =30% and the
high rate of the BT charge is 𝑆𝑂𝐶 𝑏𝑎𝑡−𝑚𝑎𝑥 =90%. Also, the low rate of
the SC charge is 𝑈 𝑠𝑐𝑚𝑖𝑛 =58 V and the high rate of the SC charge is
𝑈 𝑠𝑐𝑚𝑎𝑥 =98 V.
4.2. FLS strategy
EMS based on FLS has two main objectives. The first one is to control
the power flow and minimize the number of charging/discharging cycles
of BT to improve their lifespan. The second one is to select the exact
reference powers of BT and SC by controlling the SOC of each storage
system and make them in acceptable margins.
∙ Fuzzy logic for power flow control
The power flow control by fuzzy logic (Fuzzy toolbox ‘‘Power flow
control’’) includes three inputs and two outputs, as shown in Fig. 6.
The inputs are the difference between the wind generated power and
the required power (𝑃𝑠𝑡𝑜 ), the SOC of BT ‘‘𝑆𝑂𝐶 𝑏𝑎𝑡 ’’ and the SOC of SC
‘‘𝑈𝑠𝑐 ’’. The outputs are the command signal of switch 1 ‘‘CS1’’ and the
command signal of switch 3 ‘‘CS3’’.
The determination of the membership functions for the fuzzification
of the input and output variables of the energy supervisor are an
important phase of the fuzzy algorithm.
In this work, the membership functions have been chosen in an
empiric manner relying on our expertise on the system.
The membership functions for the three input variables and for the
two output variables must be defined (Fig. 7). Since the number of
fuzzy rules depends on the number of fuzzy sets of inputs, we will only
consider the sets relevant to the case study.
For SOC (𝑆𝑂𝐶 𝑏𝑎𝑡 and 𝑈𝑠𝑐 ), the membership functions consist of three
levels (‘‘L’’, ‘‘M’’, ‘‘H’’) which correspond to the three operating modes.
The ‘‘L’’ and ‘‘H’’ sets ensure the storage availability while avoiding
low and high saturations. The ‘‘M’’ set is used to compensate for the
oversubscribed power and to store the surplus of renewable production.
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Fig. 5. Control of HESS and DC bus voltage.

Fig. 6. Fuzzy logic power management supervisor structure.

This set also makes it possible to ensure the adjustment of the long-term
and short-term storage instructions.
The membership functions of the power deviation 𝑃𝑠𝑡𝑜 can be ‘‘N’’ or
‘‘P’’, where N and P respectively represent the discharging and charging
of the storage system.
For the membership functions of the command signals of switches 1
and 3 (CS1 and CS3), the choice of sets is such possible values of the
output variable are in the interval [−1, 1]. These functions are based on
two levels: N stands for negative and P stands for positive.
The used inference matrix is described by Tables 2 and 3 to deter-
mine the rules of control in order to associate the fuzzy inputs and

220
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Fig. 7. Block diagram of proposed input/output fuzzy-logic-based power flow control.

Table 2 Table 4
CS1 rules. CS2 rules.
CS1 𝑈𝑠𝑐 CS2 𝑆𝑂𝐶𝑏𝑎𝑡
L M H L M H
𝑃𝑠𝑡𝑜 P N N P 𝑃𝑏𝑎𝑡−𝐶𝑆1 P N N P
N P N N N P N N

Table 3 Table 5
CS3 rules. CS4 rules.
CS3 𝑆𝑂𝐶𝑏𝑎𝑡 CS4 𝑈𝑠𝑐
L M H L M H
𝑃𝑠𝑡𝑜 P N N P 𝑃𝑠𝑐−𝐶𝑆3 P N N P
N P N N N P N N

outputs. In fact, Table 2 presents the rules between 𝑃𝑠𝑡𝑜 and 𝑈𝑠𝑐 to 5. Fuzzy logic droop control technique
determine the control signal of switch 1, and Table 3 presents the rules
between 𝑃𝑠𝑡𝑜 and 𝑆𝑂𝐶 𝑏𝑎𝑡 to determine the control signal of switch 3.
∙ Selection of the exact reference power of BT and SC In this section, based on the power management supervisor outputs,
A fuzzy logic used to select the exact reference powers of BT and adaptive FLDC is proposed, which is able to inject or absorb active and
SC (Fuzzy toolbox ‘‘BT and SC control’’) includes four inputs and two reactive power ‘‘ 𝑃𝑓 and 𝑄𝑣 ’’ into the grid with high accuracy and ensure
outputs, as shown in Fig. 6. The inputs (Fig. 8) are the BT power a continuous supply of load.
(𝑃𝑏𝑎𝑡−𝐶𝑆1 ), the SOC of BT (𝑆𝑂𝐶 𝑏𝑎𝑡 ), the SOC of SC (𝑈𝑠𝑐 ), and the In the inverter output, the active and reactive powers transferred to
SC power (𝑃𝑠𝑐−𝐶𝑆3 ). The outputs (Fig. 8) are the command signals of
the load and the grid can be expressed as follows (Vasquez, Guerrero,
switches 2 and 4 (CS2 and CS4) to determine the exact reference power
Luna, Rodriguez, & Teodorescu, 2009):
of SC and BT while respecting these limits of charging and discharging.
The purpose of this fuzzy-logic toolbox is to maintain the SOC of SC and 𝑉𝑐2 ( ( ) )
BT at acceptable levels. 𝑃 = 𝑅𝑔 𝑉𝑐 − 𝑉𝑝𝑐𝑐 cos (𝛿) + 𝑋𝑉𝑝𝑐𝑐 sin (𝛿)
𝑅2𝑔 + 𝑋 2
The membership functions of 𝑃𝑏𝑎𝑡-CS1 and 𝑃𝑠𝑐-CS3 are based on two (16)
levels, N and P, to accommodate the needs of the proposed strategy, 𝑉𝑐2 ( ( ) )
𝑄= 𝑋 𝑉𝑐 − 𝑉𝑝𝑐𝑐 cos (𝛿) − 𝑅𝑉𝑝𝑐𝑐 sin (𝛿)
where N and P respectively represent the discharge and charge of each 𝑅2𝑔 + 𝑋 2
storage system. The output membership functions present the command
of switch 2 by the CS2 command signal and the command of switch 4 The powers passing through the line depend on the reactance of the
by the CS4 command signal. These functions are based on two levels: N line, the voltage levels and the voltage angles. After using the orthogonal
stands for negative and P stands for positive. linear rotational T transformation matrix, the active and reactive powers
The used inference matrix is described by Tables 4 and 5 to deter- become:
mine the rules of control to associate the fuzzy inputs and outputs. ( ′) ( )
𝑃 𝑃
Actually, Table 4 presents the rules between 𝑃𝑏𝑎𝑡-CS1 and 𝑆𝑂𝐶 𝑏𝑎𝑡 to ′ =𝑇 (17)
𝑄 𝑄
determine the control signal of switch 2, and Table 5 presents the rules
between 𝑃𝑠𝑐-CS3 and 𝑈𝑠𝑐 to determine the control signal of switch 4. with:
In this paper, the Mandani method is chosen for defuzzification using
the center of gravity (Arun & Mohan, 2017; Nikita & Garg, 2017). It is ⎛𝑋 −
𝑅⎞
⎜ 𝑍⎟
𝑇 ⎜𝑍
𝑋 ⎟⎟
intuitive, it has widespread acceptance and it is well suited to human
inputs. ⎜𝑅
⎝𝑍 𝑍 ⎠

221
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Fig. 8. Block diagram of proposed input/output fuzzy-logic-based SOC control.

GDC is characterized by a slow dynamics response, a low frequency

𝑉𝑐 𝑉𝑝𝑐𝑐 and voltage regulation, and a poor performance with renewable energy
𝑃′ = sin (𝛿) generators. For this framework, an adaptive fuzzy logic is put forward,
𝑍 (18)
𝑉 ( ) as given in Fig. 10.
𝑃 ′ = 𝑐 𝑉𝑐 − 𝑉𝑝𝑐𝑐 sin (𝛿) In order to adapt inverter output powers ‘‘ P and Q ’’ with reference
𝑍
powers ‘‘ 𝑃 ∗ and 𝑄∗ ’’, transferred to the load and the grid, the following
where 𝛿 is the transport angle, which is the phase difference between
droop control technique is defined by using transformation (17) as
the inverter output voltage and the grid voltage. As a consequence,
follows:
if 𝛿 is small, we will obtain, thanks to the limited development, the ( )
1 𝑍 ′
following relationship between the phase and the active power, as well 𝛥𝑓 = 𝑃 − 𝑃 ∗ + 𝑘𝑅 𝑘𝑣 𝛥𝑉 + 𝑘𝑅 𝑄∗
as a relationship between the reactive power and the voltage, which 𝑘𝑓 𝑋
( ) (21)
represents a foundation of the droop control as follows: 1 𝑍 ′
𝛥𝑉 = 𝑄 − 𝑄∗ + 𝑘𝑅 𝑘𝑓 𝛥𝑓 + 𝑘𝑅 𝑃 ∗
𝑉𝑐 𝑉𝑝𝑐𝑐 𝑘𝑣 𝑋
𝑃′ = 𝛿 where 𝑃 ∗ =P 𝐿 +P 𝑓 and 𝑄∗ =Q 𝐿 +Q 𝑣 are the references of the active
𝑍 (19)
𝑉 ( ) and reactive powers transferred to the load and the grid, 𝑃𝐿 and 𝑄𝐿 are
𝑃 ′ = 𝑐 𝑉𝑐 − 𝑉𝑝𝑐𝑐 the active and reactive powers demanded by the load, and 𝑃𝑓 and 𝑄𝑣
𝑍
In order to show the performance of the proposed FLDC, a compar- are the active and reactive powers transferred to the grid defined by the
ative study with conventional droop control is suggested. This GDC is following relationships:
( )
depicted in Fig. 9. 𝑃𝑓 = 𝑘𝑓 𝛥𝑓 − 𝑠𝑖𝑔𝑛 (𝛥𝑓 ) 𝛥𝑓𝑎𝑐𝑐
( ) (22)
The GDC output is defined by the following expressions (Saleh, 𝑄𝑣 = 𝑘𝑣 𝛥𝑉 − 𝑠𝑖𝑔𝑛 (𝛥𝑉 ) 𝛥𝑉𝑎𝑐𝑐
Shoresh, & Hassan, 2015):
( ) Where 𝑘𝑓 and 𝑘𝑣 are adjustable gains, 𝛥𝑓 = 𝑓 𝑝𝑐𝑐 −𝑓 𝑛 is the difference
1 𝑍 ′ between the grid and nominal frequencies, 𝛥𝑉 = 𝑉 𝑝𝑐𝑐 −𝑉 𝑛 is the
𝛥𝑓 = 𝑃 − 𝑃 ∗ + 𝑘𝑅 𝑘𝑣 𝛥𝑉 + 𝑘𝑅 𝑄∗
𝑘𝑓 𝑋 difference between the grid and nominal voltages, 𝛥𝑓 𝑎𝑐𝑐 =±0.1 Hz is the
( ) (20)
1 𝑍 ′ acceptable frequency level, and 𝛥𝑉 𝑎𝑐𝑐 =±5 V is the acceptable voltage
𝛥𝑉 = 𝑄 − 𝑄∗ + 𝑘𝑅 𝑘𝑓 𝛥𝑓 + 𝑘𝑅 𝑃 ∗
𝑘𝑣 𝑋 level.
where 𝑘𝑓 , 𝑘𝑣 , and 𝑘𝑅 are adjustable gains. The target behind using fuzzy logic in this section is to generate the
To ensure a smooth transition between grid-connected and islanded adjustment variations, which will be entrenched within the phase and
modes, we use automatic switches (𝑆5 , 𝑆6 , 𝑘1 and 𝑘2 ) as follows: amplitude of the grid voltage, to make them in better margins than those
mentioned in the standards. FLDC used in this study includes four inputs
• In the grid connected mode; automatic switches 𝑆6 and 𝑆5 are and two outputs as shown in Fig. 11.
′ ′
respectively in states 𝑉𝑝𝑐𝑐 and 𝜔𝑝𝑐𝑐 t, and 𝑘1 and 𝑘2 are closed. In The inputs are errors ‘‘ 𝑒𝑝 = 𝑃 ∗ −𝑃 ′ and 𝑒𝑞 =Q ∗ −𝑄′ ’’ and error
this mode, GDC allows maintaining the phase and the rms value of derivatives 𝑑𝑒𝑝 and 𝑑𝑒𝑞 of the active and reactive powers. The outputs
the grid voltage, at each moment, within a desirable margin more are the adjustment variations in the phase and amplitude of the grid
demanding than that mentioned in the standards. output voltage.
• In the islanding mode; AG passes to the islanded mode. Automatic The first stage of the design makes it possible to better define the
switches 𝑘1 and 𝑘2 are opened, and 𝑆6 and 𝑆5 are respectively characteristics of the variables. It is now necessary to completely define
changed to states 𝑉𝑛 and 𝜔𝑛 t. In this mode, the objective of GDC is the fuzzy subsets, i.e. to clarify their membership functions. We find
to stop any power exchange with the grid, and to impose 𝜔𝑛 and that the membership functions of the errors have a symmetrical form
𝑉𝑛 , as the phase and the voltage of the reference at the load, to creating a concentration around zero, which improves the accuracy
protect them against strong fluctuations of the grid voltage and near the point of a desired operation. However, we introduce nine
frequency. additional subsets given the sensitivity of the variables to be controlled:

222
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Fig. 9. Conventional droop control technique.

Table 6 Table 7
Rules of frequency control. Rules of voltage control.
𝛥𝜔𝑡 𝑒𝑝 𝛥V 𝑒𝑞
NBB NB NM NS Z PS PM PB PBB NBB NB NM NS Z PS PM PB PBB
N PS PS PS PS PS PS PS PS PS N PS PS PS PS PS PS PS PS PS
𝑑𝑒𝑝 M PBB PB PM PS Z NS NM NB NBB 𝑑𝑒𝑞 M PBB PB PM PS Z NS NM NB NBB
P NS NS NS NS NS NS NS NS NS P NS NS NS NS NS NS NS NS NS

NBB: Negative Big Big, NB: Negative Big, NM: Negative Medium, NS:
Negative Small, Z: Zero, PBB: Positive Big Big, PB: Positive Big, PM:
Positive Medium, and PS: Positive Small.
For the same reason, the forms of the membership functions of the
output variables are also symmetrical. However, we introduce nine
fuzzy subsets.
The membership functions of the error derivatives are chosen to
specify the system response. Nevertheless, we introduce three fuzzy
subsets: P: Positive, M: Medium, and N: Negative.
The next step is the development of a droop controller rule base. By
describing step by step the behavior of the process and the action of
the variation in the command to apply, we deduce tables (basic fuzzy
controller tables) which actually correspond to rules. The responses of
the frequency and voltage control are summarized in Tables 6 and 7.
This set of rules groups together all the possible situations of the
system evaluated by the different values attributed to ‘‘ 𝑒𝑖 ’’ and its
variation ‘‘ 𝑑𝑒𝑖 ’’ and all the corresponding values of the command
variation, where 𝑖 = 𝑝, 𝑞.
In this work, FLDC is used to exchange active and reactive powers
from the three-phase voltage source inverter to the grid by automatically
adjusting the amplitude and frequency of the output voltage in both
standalone and grid-connected modes. Furthermore, it allows having a
smooth transition between different operation modes using automatic
switches (S5, S6, S7 and S8). They are controlled by fuzzy logic.
∙ Switches S5 and S6 permit selecting between 𝑉𝑛 and 𝑉𝑝𝑐𝑐 , and 𝑓𝑛
and 𝑓𝑝𝑐𝑐 , respectively.
If 𝐴𝑖𝑠𝑙 is positive then S5 and S6 are in states 𝑉𝑝𝑐𝑐 and 𝑓𝑝𝑐𝑐 ,
respectively. Else, S5 and S6 are in states 𝑉𝑛 and 𝑓𝑛 , respectively.
∙ Switch S7 enables choosing between 0 and 𝑅2 .
If 𝐴𝑣 is positive then 𝑄𝑣 = R2 ; else, 𝑄𝑣 =0.
∙ Switch S8 allows choosing between 0 and 𝑅1 .
If 𝐴𝑓 is positive then 𝑃𝑓 = R1 ; else, 𝑃𝑓 =0.
These switches are controlled by control signals ‘‘ 𝐴𝑖𝑠𝑙 , 𝐴𝑣 and 𝐴𝑓 ’’
generated by a fuzzy logic toolbox named ‘‘Fuzzy Logic Islanding
Detection (FLID)’’. The input and output membership function of this
FLID is to ensure the transition between different operation modes, as
provided in Fig. 12.
The inputs are grid voltage 𝑉𝑝𝑐𝑐 and grid frequency 𝑓𝑝𝑐𝑐 . The outputs
are the control signals of automatic switches ‘‘ 𝐴𝑖𝑠𝑙 , 𝐴𝑣 and 𝐴𝑓 ’’. We find
that the membership functions of the inputs have a symmetrical form.
However, we introduce five fuzzy sets for each input: LB: Low Big, L:
Low, M: Medium, H: High, and HB: High Big.
For the same reason, the forms of the membership functions of the
output variables are also symmetrical. Yet, we introduce two fuzzy sets:
N stands for negative and P stands for positive.

223
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Fig. 10. Adaptive fuzzy logic droop control technique.

Fig. 11. Block diagram of proposed inputs/outputs of FLDC.

The fuzzy logic rules used for the islanding detection are obtained associate the fuzzy inputs to the fuzzy outputs are made up of three
from the analysis of the grid fluctuations. The control roles can help
AG to improve the performances of grid stability. The control rules that parts, which are as follows:

224
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Fig. 12. Block diagram of proposed FLID inputs/outputs.

Table 8 - In the grid connected mode, automatic switches S5 and S6 are respec-
𝐴𝑖𝑠𝑙 rules. tively in states 𝑉𝑝𝑐𝑐 and 𝜔𝑝𝑐𝑐 t, and S7 and S8 are respectively in
𝐴𝑖𝑠𝑙 𝑉𝑝𝑐𝑐 states 𝑅2 and 𝑅1 . The adaptive FLDC system reduces the voltage
HB H M L LB and frequency fluctuations by adopting the power supplied by
HB N N N N N the inverter to total powers ‘‘ 𝑃𝐿 +𝑃 𝑓 ’’ and ‘‘ 𝑄𝐿 +𝑄𝑣 ’’ consumed
H N P P P N by the load and the grid. It controls exported or imported powers
𝑓𝑝𝑐𝑐 M N P P P N (𝑃𝑓 and 𝑄𝑣 ) with the grid ensuring its stability.
L N P P P N
LB N N N N N
- In the standalone mode, once FLID detects an isolating condition, RDG
will pass to the islanded mode and the additional injection of ‘‘
Table 9 𝑃𝑓 and 𝑄𝑣 ’’ will have no sense, so they will be zero in this mode.
𝐴𝑣 rules. Automatic switches S7 and S8 are changed to state 0, and S5 and
𝑉𝑝𝑐𝑐 S6 are respectively changed to states 𝑉𝑛 and 𝜔𝑛 t.
HB H M L LB
𝐴𝑖𝑣 N P N P N
6. Simulations and interpretations

In this section, we will present the simulations of the developed

Table 10
𝐴𝑓 rules.
𝑉𝑝𝑐𝑐
HB H M L LB
𝐴𝑓 N P N P
• Table 8 shows the fuzzy rules used between 𝑉𝑝𝑐𝑐 and 𝑓𝑝𝑐𝑐 to obtain
command signal 𝐴𝑖𝑠𝑙 of switches 5 and 6. These rules are divided
into twenty rules.
• Table 9 presents the fuzzy rules used for 𝑉𝑝𝑐𝑐 to obtain command
signal 𝐴𝑣 of switch 7. These rules are divided into five rules.
• Table 10 gives the fuzzy rules used for 𝑓𝑝𝑐𝑐 to obtain command
signal 𝐴𝑓 of switch 8. These rules are divided into five rules.
We are conscious of the stability problematic. In our previous work
Krim et al. (2017); Krim, Abbes, Krim, and Mimouni (2018) nonlinear
methods have been used to analyze the stability of controllers, such as
Popov, the first harmonic and the Lyapunov function, as we used the
sliding mode control. However, in this work, as the Mandani fuzzy logic
method based on system knowledge is used without a mathematical
model, stability analysis is a priori difficult. It is rather proven by
simulation results. In fact, disturbances are well handled.
In addition, FLDC is used to exchange the active and reactive powers
from the inverter to the grid by automatically improving the rms
value and frequency of the load voltage in both islanded and grid-
connected modes. Besides, FLID enables having a smooth transition
between different operation modes using automatic switches (S5, S6,
S7 and S8).
control strategies to ensure the electrical distribution network stability
in the presence of the wind, load, and grid voltage and frequency
fluctuations. The parameters used in this study are summarized in
Table 11:
N The simulation conditions presented in Fig. 13 (variations in wind
speed, grid frequency, grid voltage and load active power) evaluate the
control strategy to test their feasibility in all operation modes. They are
discussed for different scenarios to see the reaction of grid stability.
Wind speed is close to reality and varies between 7 and 13 m/s, as
shown in Fig. 13(a) (Edmund, 2004).
The profile of the active power demanded by the load is based on the
actual consumption patterns of a typical family of four people in Tunisia
(Randa, Ghada, & Lotfi, 2015), as provided in Fig. 13(b).
Fig. 13(c) and Fig. 13(d) respectively depict the grid voltage and
frequency fluctuations at PCC. These fluctuations are presented to test
the feasibility of the proposed control to detect the grid stability state,
the islanding mode in case of a grid fault and the grid-connected mode
when voltage and frequency fluctuations do not exceed their maximal
margins mentioned in the standards (Mouna, Achraf, & Lotfi, 2015;
Standard EN 50160, 2004).
6.1. Fuzzy logic power supervision
According to the internal specifications of both storage devices
developed in this study, a low pass filter is used with constant time
𝜏 = 6 s. Fig. 14 illustrates the evolution of the storage power to satisfy
the production scenario.
Figs. 15 and 16 show the response shares of the two used storage
technologies. As predicted in the design phase, BT ensures long stor-
age regimes to ameliorate its lifespan by minimizing the number of

225
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Fig. 13. Simulation conditions: (a) Wind speed, (b) Active power demanded by load, (c) Grid voltage, (d) Grid frequency.

Table 11 demanded ones (Fig. 16). This simulation test is executed with the
Parameters of studied system. classic control in MATLAB/SIMULINK and with FLS. Fig. 15(a) and
Parameters Value 15(b) represent the reference BT storage power deduced with classical
PMSG Stator resistance 𝑅𝑠 0.82 Ω control and fuzzy logic, respectively. Similarly for SC, Fig. 16(a) and
Stator inductance 𝐿𝑠 0.0151 mH 16(b) present the reference SC storage power deduced with classical
Magnet flux 𝛷𝑚 0.4832 Wb
control and fuzzy logic, respectively.
Nominal power 3.5 kW
Number of pole pairs p 4 The low rate of the charge of BT is 30% and the high rate of the
Inertia J 99*10−4 kg m2 charge of BT is 90%. Also, the low rate of the charge of SC is 58 V and
Friction f 10−3 N m s rad−1 the high rate of the charge of SC is 98 V. Figs. 15(c) and 16(c) show the
Wind turbine Nominal power 3.5 kW BT and SC currents, respectively. BT reacts more slowly to the needs.
Blade radius R 2m
Contrariwise, SC provides the transient currents. When SC reaches its
Optimal tip speed ratio 𝜆𝑜𝑝𝑡 8.15
Maximum power coefficient 𝐶𝑝𝑚𝑎𝑥 0.4794 low rate of charge (𝑈𝑠𝑐𝑚𝑖𝑛 ), the totality of lack will be provided by BT.
Density of air 𝜌 1.225 Kg m3 Similarly, when BT reaches its low rate of charge, the totality of lack
Battery 𝑆𝑂𝐶 𝑚𝑎𝑥 0.9 will be provided by SC, in order to protect them from discharge excess.
𝑆𝑂𝐶 𝑚𝑖𝑛 0.3 When SC reaches 98 V, its charge will be stopped and the production
Internal resistance 𝑅𝑖 0.15 Ω
excess must be absorbed by BT. Likewise, when BT reaches 90%, its
Nominal voltage 𝐸𝑏𝑎𝑡0 60 V
Battery inductor filter 𝐿𝑏𝑎𝑡 10−6 H charge will be stopped and the production excess must be absorbed
Supercapacitor Capacitance 𝐶𝑠𝑐𝑐 94 F+20%–0% by SC, in order to protect them from overcharge. Thus, the simulation
Rated voltage 78 V results demonstrate the efficiency of proposed FLS by maintaining 𝑈𝑠𝑐
DC maximum current 50 A (Fig. 15(d)) and 𝑆𝑂𝐶 𝑏𝑎𝑡 (Fig. 16(d)) at acceptable levels.
SC inductor filter 𝐿𝑠𝑐 10−6 H
By comparing the classical and fuzzy-logic simulation tests, we
Equivalent series resistance 𝑅𝑠𝑐𝑐 12.5 mΩ
[𝑈𝑠𝑐𝑚𝑖𝑛 , 𝑈𝑠𝑐𝑚𝑎𝑥 ] [58 V, 98 V]
notice that EMS, using FLS, gives excellent results.
Leakage current 0.15 A, 75 h, 25 ◦ C. Consequently, the proposed fuzzy logic energy management supervi-
Operating temperature −40 ◦ C to +65 ◦ C sor represents a reliable and efficient energy management. However, the
DC bus Capacitance C 2200 μF simulation results prove the effectiveness of the suggested strategy by
DC voltage 400 V
ensuring the balance between production and consumption (Fig. 17(a))
RLC filter Filter resistance 𝑅𝑓 0.2 Ω
Filter inductance 𝐿𝑓 20 mH
to keep a DC bus voltage stable at 400 V (Fig. 17(b)). In the proposed
Filter capacitance 𝐶𝑓 10 μF simulations, BT and SC are able to respond to demands of the grid and
the load. During the transitions of the wind power, the load and the grid,
SC replies directly to the need of the load and the grid by providing or
absorbing peak currents. BT reacts more slowly to the needs knowing
that SC provides the transient currents (Fig. 17(c)).
However, for the overproduction of a wind generator, and if the
HESS reaches its high rate of charge, the wind turbine will operate in
a limited mode to generate only the power quantity demanded by the
load and the grid. When HESS reaches its low rate of charge and the
wind power is unable to meet the demands, a no-priority load will be
disconnected to ensure the balance between produced and consumed
powers. Then the real, produced, stored and demanded powers are
depicted in Fig. 17(a).
Fig. 14. Stored power.

6.2. Fuzzy logic droop control technique

charge/discharge cycles (Fig. 15). Whereas, SC responds to the rapid According to the grid frequency and the voltage state, the active
changes in the difference between the produced wind power and the generator can operate in grid-connected and standalone modes (FLID).

226
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Fig. 15. BT storage system, (a) BT reference power with FLS, (b) BT reference power without FLS, (c) BT current, (d) BT SOC.

Fig. 16. SC storage system, (a) SC reference power with FLS, (b) SC reference power without FLS, (c) SC current, (d) SC SOC.

Fig. 17. (a) Simulated power assessment, (b) DC-bus voltage, (c) Simulation of 𝑃sc * and 𝑃bat * with variable demanded power.

Fig. 18(c) shows the periods in which AG operates in a separated
‘‘𝐴𝑖𝑠𝑙 = −0.5 ’’ or connected ‘‘ 𝐴𝑖𝑠𝑙 = 0.5 ’’ state. In a grid-connected
mode, the quantities of active and reactive powers are exchanged with
the grid to reduce the fluctuations in the voltage and the frequency, and
to make them at more demanding margins than those mentioned in the
standards. These quantities are depicted in Fig. 18(a) and Fig. 18(b),
respectively.

227
Y. Krim et al.
Fig. 18. (a) Active power exchanged with grid, (b) Reactive power exchanged with grid,
(a) Signal of operation mode.
In this part, we present a comparative study to facilitate choosing
between GDC and FLDC.
Fig. 19 presents the grid frequency before and after regulation with
fuzzy and generalized droop control. According to the latter, the impact
of active power ‘‘ 𝑃𝑓 ’’ exchanged with the grid on the reduction in
the frequency fluctuation is remarkable. These variations are well mini-
mized, which demonstrates the importance of the renewable generator.
Similarly, with a grid voltage, the impact of reactive power ‘‘ 𝑄𝑣 ’’
exchanged with the grid on the reduction in the voltage fluctuation is
remarkable (Fig. 20). With elaborate FLDC, it is remarkable that the
voltage and frequency levels are well minimized and become in optimal
areas ([50.1, 49.9] and [235, 225], respectively).
The frequency and voltage errors are given in Figs. 21 and 22. We
find that with FLDC, the voltage and frequency levels are well minimized
Fig. 19. Grid frequency.
Fig. 20. Grid Voltage.
228
Control Engineering Practice 81 (2018) 215–230
with negligible errors. Comparing the two control techniques in terms
of complexity, FLDC is much simpler to synthesize than the classic
droop control. From a robustness point of view, the distributed electrical
network becomes more stable with the proposed FLDC.
The results show that this methodology can be successfully applied
for a grid-connected renewable distributed generator equipped with
a hybrid storage system in order to improve its energy efficiency. In
addition, the results highlight the feasibility of the proposed control
strategy of the distributed generator in different operation modes for an-
cillary services (ensuring grid stability) under grid faults. The proposed
intelligent control method allows improving the plans of variation in
the grid frequency and voltage. It makes them in optimal margins with a
significant reduction in disturbances caused by the limited performances
of the generalized control technique. The latter is characterized by a
high overshoot and an important response time. Certain comparison
criteria between fuzzy control and GDC under the aforementioned cases
of study are summarized in Table 12:
This comparative study demonstrates that the suggested intelligent
control strategy is suitable for this RDG application, showing a highly
robust behavior for the adjustment of the grid frequency and voltage at
PCC.
7. Conclusion
In this paper, an intelligent control strategy based on fuzzy logic
has been proposed, which facilitates the improvement of electrical
network stability by means of an active generator. The main purpose
of the intelligent control is to make an active generator participate
in system services by adjusting the grid voltage and frequency and
ensure a continuous supply load in case of grid failure. This control
has three control parts. The first one is the fuzzy logic supervisor,
which manages the power flow to keep the DC bus voltage constant
in order to deal with the load, the wind-power variations and the grid
fluctuations by maintaining BT and SC at admissible intervals of their
SOC. The second one is FLDC used to control the transferred active
and reactive powers with the grid ensuring its stability by ameliorating
the frequency and amplitude of its output voltage. The third part is
FLID, which ensures the transition between the grid-connected and
separated modes and calculates the necessary quantity of active and
reactive powers to be injected or absorbed by the grid to guarantee its
stability. Based on the comparative study between generalized and fuzzy
Y. Krim et al. Control Engineering Practice 81 (2018) 215–230

Fig. 21. Grid frequency error.

Fig. 22. Grid voltage error.

Table 12
Comparative study between proposed control approach and conventional control.
Comparison criteria Grid frequency f 𝑝𝑐𝑐 stability Grid voltage V 𝑝𝑐𝑐 stability
At t = 45 s At t = 60 s At t = 45 s At t = 60 s
GDC FLDC GDC FLDC GDC FLDC GDC FLDC
Percentage of overrun 6% 0.2% 5.8% 0.4% 5.21% 1.7% 10.43% 1.6%
Stabilization time (s) 0.17 0.04 0.2 0.06 0.16 0.035 0.172 0.06
Rising time (s) 0.2 0.11 0.2 0.15 0.2 0.11 0.2 0.15
Control capability Yes Yes Yes Yes
Control performance High peak Good High peak Good High peak Good High peak Good
Robustness Poor Robust Poor Robust Poor Robust Poor Robust

logic control, the enhanced voltage and frequency regulation and the
power management performance have been observed with the suggested
fuzzy logic. The simulation results have demonstrated the validity of the
proposed intelligent control strategy.
We will start the concept and experimental implementation of the
generalized control and the proposed control method for the studied
renewable distributed generator, in order to make a real-time com-
parative study between these two control techniques. In addition, in
future work, we will develop an intelligent control based on fuzzy
logic by utilizing nonlinear and asymmetrical membership functions to
ensure more power grid stability. An optimization technique will be put
forward to improve the choice of the membership functions by using
neural networks or genetic algorithms for instance.
References
Arun, N. K., & Mohan, B. M. (2017). Modeling, stability analysis, and computational
aspects of some simplest nonlinear fuzzy two-term controllers derived via center of
area/gravity defuzzification. ISA Transactions, 70, 16–29.
Cabrane, Z., Ouassaid, M., & Maaroufi, M. (2017). Battery and supercapacitor for photo-
voltaic energy storage: a fuzzy logic management. IET Renewable Power Generation,
11(8), 1157–1165.
Edmund, C. C. C. (2004). Field measurement and experimental study of wind speed profile
during thunderstorms. Journal of Wind Engineering and Industrial Aerodynamics, 92(3),
275–290.
Erdinc, O., & Uzunoglu, M. (2011). The importance of detailed data utilization on
the performance evaluation of a grid-independent hybrid renewable energy system.
International Journal of Hydrogen Energy, 36, 12664–12677.
Furtado, E. C., Aguirre, L. A., & Torres, L. A. B. (2008). UPS parallel balanced operation
without explicit estimation of reactive power-a simpler scheme. IEEE Transactions on
Circuits and Systems II: Express Briefs, 55(10), 1061–1065.
Glasnovic, Z., & Margeta, J. (2011). Vision of total renewable electricity scenario.
Renewable Sustainable Energy Review, 15(4), 1873–1884.
Guerrero, J. M., Matas, J., de Vicuna, L. G., Berbel, N., & Sosa, J. (2006). Wireless-control
strategy for parallel operation of distributed-generation inverters. IEEE Transactions
on Industrial Electronics, 53(5), 1461–1470.
Hou, G., Xing, F., Yang, Y., & Zhang, J. (2015). Virtual negative impedance droop method
for parallel inverters in microgrids. In Proceedings of the 10th conference on industrial
electronics and applications (pp. 1009–1013). IEEE.
Il-Yop, C., Wenxin, L., David, A. C., Emmanuel, G. C., & Seung-Il, M. (2010). Control
methods of inverter-interfaced distributed generators in a microgrid system. IEEE
Transactions on Industry Applications, 46, 1078–1088.
Islam, Z., Djamel, B., Abdel-Moumen, D., Bilal, A., & Ikram. El, A. (2018). Hierarchical
control for flexible microgrid based on three-phase voltage source inverters operated
in parallel. Electrical Power and Energy Systems, 95, 188–201.
Jia, H., Mu, Y., & Qi, Y. (2014). A statistical model to determine the capacity of battery–
supercapacitor hybrid energy storage system in autonomous microgrid. Electrical
Power and Energy Systems, 54, 516–524.
Juan, C. V., Josep, M. G., Alvaro, L., Pedro, R., & Remus, T. (2009). Adaptive droop control
applied to voltage-source inverters operating in grid-connected and islanded modes.
IEEE Transactions on Industrial Electronics, 56, 4088–4096.
Krim, Y., Abbes, D., Krim, S., & Mimouni, M. F. (2017). Classical vector, first-order sliding
mode and high-order sliding-mode control for a grid-connected variable speed wind
energy conversion system: a comparative study. Wind Engineering, 42(1), 16–37.
Krim, Y., Abbes, D., Krim, S., & Mimouni, M. F. (2018). Power management and second-
order sliding mode control for standalone hybrid wind energy with battery energy
storage system. Proceedings of the Institution of Mechanical Engineers, Part I (Journal of
Systems and Control Engineering), 1–23.
Lee, C., Chuang, C., Chu, C., & Cheng, P. (2009). Control strategies for distributed energy
resources interface converters in the low voltage Microgrid. In 2009 IEEE energy
conversion congress and exposition, (p. 20-22–9).
Masmoudi, A., Abdelkafi, A., & Krichen, L. (2011). Electric power generation based on
variable speed wind turbine under load disturbance. Energy, 36, 5016–5026.
Matas, J., Castilla, M., Vicuña, L. G. D., Miret, J., & Vasquez, J. C. (2010). Virtual
impedance loop for droop-controlled single-phase parallel inverters using a second-
order general-integrator scheme. IEEE Transactions on Power Electronics, 25, 2993–
3002.

229
Y. Krim et al.
Matthieu, D., Arnaud, D., Karl, S., Moe, T., Marc, M., & Richard, R. (2017). Battery energy
storage system battery durability and reliability under electric utility grid operations:
analysis of 3 years of real usage. Journal of Power Sources, 338, 65–73.
Mouna, R., Achraf, A., & Lotfi, K. (2015). A micro-grid ensuring multi-objective control
strategy of a power electrical system for quality improvement. Energy, 88, 351–363.
Nikita, G., & Garg, Rachna (2017). Tuning of asymmetrical fuzzy logic control algorithm
for SPV system connected to grid. International Journal of Hydrocarbon Engineering, 42,
16375–16385.
Petronela, P., Dhaker, A., Christophe, S., Stephane, B., Julien, P., & Benoit, R. (2016).
Multi-criteria fuzzy-logic optimized supervision for hybrid railway power substations.
Mathematics and Computers in Simulation, 130, 236–250.
Randa, K., Ghada, B., & Lotfi, K. (2015). Demand side management of household
appliances in stand-alone hybrid photovoltaic system. Renewable Energy, 81, 123–135.
Sahyoun, S., Djouadi, S. M., & Shankar, M. (2015). Optimal control of droop controlled
inverters in islanded microgrids. IFAC-Papers, 48, 363–368.
Saleh, A., Shoresh, S., & Hassan, B. (2015). A fuzzy logic-based droop control for
simultaneous voltage and frequency regulation in an AC microgrid. Electrical Power
and Energy Systems, 64, 148–155.
Shuai, Z., Shanglin, M., Jun, W., John, S., Wei, T., & Yan, F. (2016). Droop control method
for load share and voltage regulation in high voltage microgrids. Journal of Modern
Power System Clean Energy, 4, 76–86.
Song, Z., Hofmann, H., Li, J., Hou, J., Han, X., & Ouyang, M. (2014). Energy management
strategies comparison for electric vehicles with hybrid energy storage system. Applied
Energy, 134, 321–331.
230
Control Engineering Practice 81 (2018) 215–230
Standard EN 50160. (2004). Voltage characteristics in public distribution systems.
Tayaba, U. B., Roslana, M. A. B., Hwaia, L. J., & Kashifa, M. (2017). A review of droop
control techniques for microgrida review of droop control techniques for microgrid.
Renewable and Sustainable Energy Reviews, 76, 717–727.
Tolani, S., & Sensarma, P. (2012). An improved droop controller for parallel operation of
single-phase inverters using R-C output impedance. In Proceedings of the 2012 IEEE
international conference on power electronics, drives and energy systems (pp. 1–6).
Vasquez, J. C., Guerrero, J. M., Luna, A., Rodriguez, P., & Teodorescu, R. (2009). Adaptive
droop control applied to voltage-source inverters operating in grid-connected and
islanded modes. IEEE Transactions on Industrial Electronics, 56(10), 4088–4096.
Victor, H., Aitor, M., Haizea, G., Ion, E-O., Igor, V., & Haritza, C. (2016). Adaptive energy
management strategy and optimal sizing applied on a battery-supercapacitor based
tramway. Applied Energy, 169, 831–845.
Wang, Y., Liu, C., Pan, R., & Chen, Z. (2017). Modeling and state-of-charge prediction
of lithium-ion battery and ultracapacitor hybrids with a co-estimator. Energy, 121,
739–750.
Youssef, K., Saber, K., & Mohamed, F. M. (2017). Control and management strategy
of a generation system made by variable-speed wind turbine and super - capacitor
energy storage system for standalone Applications. International Journal on Electrical
Engineering and Informatics, 9(3), 567–584.
Ziyou, S., Jun, H., Heath, H., Jianqiu, L., & Minggao, O. (2017). Sliding-mode and
lyapunov function-based control for battery/supercapacitor hybrid energy storage
system used in electric vehicles. Energy, 122, 601–612.