Received September 29, 2020, accepted October 10, 2020, date of publication October 21, 2020, date of current

version November 2, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.3032705

Power Management Strategy Based on Adaptive

Neuro Fuzzy Inference System for AC Microgrid
HESHAM M. FEKRY 1 , AZZA AHMED ELDESOUKY 2 , AHMED M. KASSEM3 ,
AND ALMOATAZ Y. ABDELAZIZ 4 , (Senior Member, IEEE)
1 Department of Electrical Engineering, Egyptian Propylene and Polypropylene Company, Port Said 42511, Egypt
2 Department of Electrical Power, Faculty of Engineering, Port Said University, Port Fouad 42526, Egypt
3 Department of Electrical Engineering, Faculty of Engineering, Sohag University, Sohag 82524, Egypt
4 Faculty of Engineering and Technology, Future University in Egypt, Cairo 11835, Egypt

Corresponding author: Hesham M. Fekry ([email protected])

ABSTRACT Microgrids (MGs) have been widely implemented as they increase the efficiency and resiliency
of electrical networks. However, the uncertain nature of renewable energy resources (RERs) integrated into
the MGs usually results in different technical problems. System stability, the most challenging problem,
can be achieved via a robust power management strategy (PMS) of the MG. This paper introduces a PMS
based on adaptive neuro fuzzy inference system (ANFIS) for AC MG consisting of a diesel generator (DG),
a double fed induction generator (DFIG) driven by a wind turbine (WT) and a solar photovoltaic (PV)
panel. The proposed strategy aims to achieve MG power balance, decrease DG fossil fuel to minimum
consumption, keep the MG voltage stability and finally tracking the maximum power point (MPP) of
each RER. Metaheuristic optimization techniques; including genetic algorithm (GA) and particle swarm
optimization (PSO), are employed to train the ANFIS to accomplish the desired objectives and fulfill the gen-
eration/consumption balance. The proposed AC MG with the PMS is simulated by the MATLAB/Simulink
software in order to analyze the system performance under different climatic conditions. The simulation
results under symmetrical and asymmetrical electrical faults validated the effectiveness of the proposed
strategy.

INDEX TERMS Microgrid, renewable energy resources, power management strategy, voltage stability,
adaptive neuro fuzzy inference system, double fed induction generator, genetic algorithm, particle swarm
optimization.

NOMENCLATURE R rotor radius.

CP efficiency coefficient of the wind turbine. Rs DFIG stator resistance.
G solar irradiation. Rr DFIG rotor resistance.
Ig PV generator current. s slip.
Lr DFIG rotor leakage inductance. t time interval.
Ls DFIG stator leakage inductance. Vg PV generator voltage.
PDG diesel generator output active power. v wind speed.
PGSC grid side converter active power. ωr rotor angular frequency.
Pmec. diesel generator input mechanical power. ωs synchronously rotating reference frame angular
Ppv PV generator active power. velocity.
Pr DFIG rotor active power. ωslip slip angular frequency.
Ps DFIG stator active power. ωt turbine rotational speed.
Pω extracted wind power. β blade pitch angle.
QS DFIG stator reactive power. λ tip-speed ratio.
ρ air density.

I. INTRODUCTION
The associate editor coordinating the review of this manuscript and Today, as the world population is drastically increasing, the
approving it for publication was Canbing Li. excessive consumption of fossil fuels has led to higher levels
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
VOLUME 8, 2020 For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ 192087
 H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

of pollution, greenhouse gas emissions and perceptible cli- Based on the above studies, it is noticeable that wind and
matic changes that pose a dangerous threat to the anthro- PV are the two leading RERs that have come to the forefront
pological and natural ecosystems [1], [2]. These harmful of the scientific and experimental research in the field of clean
influences have highlighted the need for renewable energy renewable power generation [15], [16]. However, the inter-
resources (RERs) as safe and clean forms for power gen- mittent and stochastic natures of them have complicated their
eration [3]. Meanwhile, Microgrids (MGs) are considered use and therefore, it is better to be merged with conventional
an effective way of managing the RERs integration on both diesel generator (DG). Studies have shown that the reliability
the distribution system level and the user level [4]. The use of a hybrid wind and solar PV system is twice as much as for
of MGs allows the RERs to operate independently during either technology used alone [17]. These combined systems
main grid failure. This in turn helps efficient transformation fulfill the energy security and stability as well as increased
of the passive network into an active one, bidirectional and power quality and whole-system efficiency [18].
controlled power flow management, reliable and continu- In our study, a PMS based on ANFIS is proposed for a MG
ous supply, power quality enhancement, and clean environ- consisting of a double fed induction generator (DFIG) driven
ment [3]. They can offer an affordable and effective means by a WT and a solar PV panel as a hybrid renewable energy
of enhancing the energy supply of the continuously growing system in combination with a DG. The implementation of
communities [2], [4]. In the near future, electricity generation ANFIS controller is promoted for various reasons. The learn-
networks will need to accommodate RERs on larger scales ing ability of artificial neural network (ANN) eliminates the
and facilitate their connection with major grids at suitable basic problems in fuzzy system design, which provides the
locations [5]. parameter optimization ability with a high degree of tolerance
Undoubtedly, the well-designed power management (PM) to uncertainty and possible system disturbances. Therefore,
is essential to a functional MG as the distribution feature it can deal with any unexpected changes in the environmental
renders their control structure greatly different from that conditions even if not predefined or previously occurred.
of the traditional/central power plants. Many demonstration Moreover, it can easily correlate the inhomogeneous inputs
projects and laboratory testing systems have been designed and calculate the output without complex mathematical mod-
to verify one or more specified PM performances [6]. els. It continuously extends its base of knowledge through
In [7], the PM for a MG consisting of a photovoltaic (PV), the addition of new rules which makes the system upgrad-
a wind turbine (WT), a distributed generator, and a bat- ing easier and faster with no need for long shutdowns [19].
tery bank with linear and non-linear loads was achieved To achieve optimum MG energy operation and control, meta-
through the utilization of the adaptive neuro fuzzy infer- heuristic optimization algorithms such as the genetic algo-
ence system (ANFIS) with elephant herd optimization (EHO) rithm (GA) and particle swarm optimization (PSO) are further
algorithm. The authors in [8] introduced a PM system for proposed for training the ANFIS [20].
PV-battery-hydro based MG capable of regulating the fre- Depending on the climatic changes, different input vari-
quency and voltage of the MG and load leveling and orga- ables are collected to the ANFIS. These are the wind speed,
nized the active power circulating during battery charging and the solar irradiation and the ambient temperature. The output
discharging. Similarly, in [9], a PV/battery hybrid distributed is the mechanical power of the DG to optimally dispatch the
power generation system was proposed with a power manage- required generated power to overcome the uncertainty of the
ment strategy (PMS) of only one integrated three-port power RERs. Different case studies depending on the weather con-
converter. In [10], an efficient PMS for a small-scale hybrid ditions and the state of the network are studied. In this work,
wind-solar-battery RER based MG was designed to maintain the PV is used to drive the rotor of the DFIG with different
the MG power balance and accommodate the different varia- scenarios depending on the density of the solar irradiation.
tions in the RERs and the load demand. In [11], an improved The main contributions of this work can be summarized
PMS of a hybrid direct current DC MG system consisting of as:
a PV cell, WT generator, battery energy storage, fuel cell and 1- A smart PMS for a MG that blends renewable and
an electrolyzer was adopted. The transient response of the traditional resources based on ANFIS is proposed.
voltage was improved using Takagi–Sugeno fuzzy control. 2- GA and PSO as metaheuristic optimization algorithms
A MG including diesel, battery storage, and solar plants are presented for ANFIS’s parameters enhancement to
was proposed in [12]. An effective PM was achieved through improve the system performance.
a time domain signal processing approach to filter the fluc- 3- The climatic parameters are the inputs of the proposed
tuating output power of the solar plant. The study in [13] ANFIS to deduct the most accurate mechanical power
presented a PMS of a MG based on a modified Tribe- particle of the DG, making the system more stable and efficient.
swarm optimization (PSO), a demand response program and a 4- The DFIG is often operating on sub-synchronous mode
power scheduling approach. In [14], the authors used a model in this work, promoting the connection of the PV ter-
predictive control theory as the mathematical framework and minals to DC link to save the use of a separate inverter
a fuzzy prediction interval model as the prediction model rep- between the PV and the MG load bus which, in turn,
resenting both nonlinear dynamic behavior and uncertainty in saves money and ensures that the resulting voltage and
the RERs. currents are smoothed.

192088 VOLUME 8, 2020

H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

5- Different case studies are performed considering the they are achieved through a control strategy that predicts
different climatic and temporary fault conditions in the generated power output of the renewable sources and
order to validate the effectiveness of the proposed accordingly calculates the needed mechanical power input.
system control. This accurately determines the minimum amount of the diesel
Accordingly, this paper is arranged as follows: In section engine fuel required to complete the two megawatt of the
II, the system used is described including the description mentioned residential load and to achieve the whole-system
of the proposed ANFIS structure with different optimiza- power balance.
tions algorithms and the model of each energy resource. In general, the ANFIS is an intelligent system based
Section III presents the simulation results and discussion. on learning and combines data processing ability of ANN
Finally, the conclusions are introduced in section IV. and inference capability of Takagi-Sugeno fuzzy inference
system. The ANFIS controller uses the ‘‘IF–THEN’’ rules
according to the learning algorithms to decrease the output
error based on the training data sets [21], [22].
The main ANFIS structure consists of five functional
blocks (rule base, database, a decision making unit, a fuzzyfi-
cation and a defuzzyfication interface modules). The classic
hybrid learning algorithm (HLA) for ANFIS uses a combi-
nation of gradient descent (GD) and least squares estima-
tor (LSE) in a two-pass parameter update process [23].
The design of the proposed ANFIS comprises three inputs
and one output. The first input is the wind speed represent-
ing the output power generated from the DFIG driven by
WT. The second and the third inputs are the solar irradia-
tion and ambient temperature as indications of the output
power extracted from PV generator. The ANFIS output is
the expected mechanical power that should be dispatched to
DG to complete the power required to balance the residential
load as previously illustrated. 385 training samples are used
FIGURE 1. The schematic diagram of the proposed hybrid power system to train the proposed ANFIS, GA-ANFIS and PSO-ANFIS
in MG. controllers. Table 1 presents ten random training samples.
The internal structure of the proposed ANFIS controller is
II. ISOLATED MG DESCRIPTION displayed in Fig. 2. Each input has a triangular membership
The schematic diagram of the proposed MG is shown in Fig.1. number of three while the number of rules is twenty seven.
The MG is formed of three different power sources, two of
them are renewable sources (DFIG driven by a WT and PV)
and the third source is a traditional DG. The MG is respon- TABLE 1. Ten random samples from ANFIS training samples.
sible for continuously supplying a two-megawatt residen-
tial load. The DFIG stator is directly connected to the grid
whereas the PV is directly connected to a DC bus through a
maximum power point (MPP) boost circuit to feed the DFIG
rotor through the rotor side converter (RSC) and also supply
the grid occasionally through grid side converter (GSC) via
a chock coil. Meanwhile, the DG stator is coupled with the
mentioned grid through a step up power transformer that
raises the voltage from 400 V to 575 V as the main voltage of
MG and load.
The PMS of the proposed system should be capable of
executing the following objectives:
1. Achieving the MG voltage stability.
2. Keeping power balance in the MG.
3. Decreasing the fuel consumption of the DG to a mini-
mum value. Additionally, the results of the ANFIS are further com-
4. Tracking the MPP of wind and PV resources. pared to those of (GA-ANFIS), and (PSO-ANFIS). The valid-
The first and fourth objectives are carried out by the own ity of the three ANFIS controllers is verified by the simulation
design control for each of DFIG driven by WT, PV, and DG as results carried out using MATLAB/ Simulink environment.
briefly discussed later on. For the second and third objectives, Table 2 displays the applied parameters of the GA and PSO.

VOLUME 8, 2020 192089

 H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

FIGURE 2. The internal structure of the proposed ANFIS controller trained

by HLA algorithm.

TABLE 2. The applied parameters for the proposed GA and PSO.

FIGURE 3. The main power flow segments of the proposed hybrid power
system during most of the day: (a) PV supplies both of the DFIG rotor
power and the MG load directly during high solar irradiation. (b) PV
supplies the DFIG rotor power only when the sunlight is light. (c) The
DFIG rotor power is supplied by the PV and the MG when the solar
irradiation becomes weaker. (d) The DFIG rotor power supplied only by
the MG during solar irradiation fading.

For the proposed system, the torque speed characteristic the following relation:
of the WT, as will be displayed later, drives the DFIG to
run always in sub-synchronous operation mode even in the Pr = sPs (1)
case of high or low wind speeds that may be expected in the where Pr and Ps are the DFIG rotor and stator power, respec-
proposed geographical site of installation. This can be quite tively and s is the slip.
advantageous as all the mechanical power output of the WT When the solar irradiation declines, the PV power is
along with any electrical power received from the rotor will decreased and feeds the rotor only. In case that the sunlight
be exploited by the DFIG stator. becomes even feebler, the PV cannot even feed the rotor alone
The hybrid power system concept is specifically applied and the GSC acts as a converter contributing to the rotor
in this work to avoid any sudden disturbances in the MG supply with an amount of power from the MG through the DC
resulting from probable failure in the renewable components. bus. Finally, when the solar irradiation fades away, the rotor
The DG is always in service supplying the grid and the precise is totally fed by the MG.
amount of its contribution is determined according to the
proposed ANFIS controller as previously discussed. A. THE PV
Fig. 3 summarizes the main power flow segments of the The PV generator has a nonlinear voltage current characteris-
proposed hybrid power system during most of the day. During tic based on solar irradiation and ambient temperature. It can
the maximum solar irradiation, a part of the PV power output be expressed from the following formula:
supplies the DFIG rotor and the rest is fed to the grid through  
1 GIphg − Ig + Iog
the GSC which acts as an inverter. The power rating of the Vg = ln − Ig Rsg (2)
used PV array at the maximum expected solar irradiation and Ag Iog
minimum expected ambient temperature is chosen to be more where Vg & Ig are the PV generator voltage and current,
than the highest DFIG rotor power consumption according to respectively; Ag = 3/Ns is the PV generator constant;

192090 VOLUME 8, 2020

H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

FIGURE 4. PV characteristics. (a) The current and power versus PV voltage characteristics with solar irradiation changing. (b) The current and power
versus PV voltage characteristics with ambient temperature changing.

3 =q/(ε×Z×U), is the solar cell constant; q= 1.602 × TABLE 3. The main technical specifications of the used PV array at
1000W/m2 solar irradiation, 25 c◦ ambient temperature.
10−19 C. is the electron charge; Z= 1.38 × 10−23 J/K is
Boltzman constant; U= 298.15; K is the absolute tempera-
ture; ε =1.1 is the completion factor; Rsg = Rs × Ns Np is


the PV generator series resistance; Ns is the series connected

solar cells; Np is the parallel paths; Rs is the series resistance
per cell; Iphg = Iph × Np is the insolation-dependent photo
current of the PV generator; Iph = 4.8 A is the photo current
per cell; Iog = Io × Np is the PV generator reverse saturation
current; Io =2.58e−5 A is the reverse saturation current per
cell; G is the solar insolation in per unit, and 1.0 per unit of
G=1000 W/m2 [24].
Fig. 4 displays the characteristics of the drawn current strings and each string consists of 5 series modules with
and output power versus the PV terminal voltage along the attached MPP tracking boost circuit. The philosophy of PV
changing of both the solar irradiation and the ambient tem- power rating selection, as displayed in the proposed array
perature for the used PV panel. The output power and the technical specifications, Table 3, is to determine the sufficient
short circuit current increase proportionally as the solar irra- power that should supply the DFIG for optimum running at
diation increases whereas they decline as the ambient tem- 1000 W/m2 solar irradiation and 25 c◦ ambient temperature
perature decreases and vice versa. The characteristics clearly during all expected wind speeds and in case of any excess
indicate that there is a unique point at which the PV can power, it feeds the grid directly.
produce the maximum power at each certain solar irradiation
and temperature. Many methods are implemented to track B. THE DFIG DRIVEN BY WT
this maximum power operating point by changing the pulse The wind energy conversion system transforms the wind
width modulation duty cycle of a boost or buck boost circuit kinetic energy into mechanical energy by means of the aero-
installed on the terminal of PV panel. In this work, the tra- dynamic forces that produce lift on the rotor shaft blades.
ditional P&O algorithm is used. This algorithm depends on It can be calculated as [26]:
the comparison between the recent PV power output (Pn ) and 1
the previous (Pn−1 ) until the difference between them,1P, Pω = ρπ R2 v3 CP (λ, β) (3)
2
reaches zero. This method has the drawback of the slow
response to the irradiation and temperature changes, however where Pω is the extracted wind power, ρ is the air density
it is still the simplest method [25]. The proposed PV array (1.225 kg/m3 ), R is the rotor radius, v is the wind speed and
type is SunPower SPR 305E-WHT-D built from 440 parallel CP is the efficiency coefficient of the turbine which is the

VOLUME 8, 2020 192091

 H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

function of the tip-speed ratio λ and the blade pitch angle β. TABLE 4. The main technical specifications of WT and DFIG.
The tip-speed ratio λ is expressed as:

ωt R
λ= (4)
v

where ωt is the turbine rotational speed.

The DFIG is the most commonly used WT-driven gen-
erator especially when high power ratings are needed as
intended in the proposed system. This technology has been
widely implemented due to its numerous advantages. The
DFIG-based WTs are able to capture more wind energy
and support the grid with large reactive power during short
disturbance events. However, DFIG is very sensitive to the
possible grid faults [27], [28]. It has three modes of opera-
tion categorized according to the operating rotational speed
and the rotor power direction. In the ‘‘super-synchronous’’
mode, the operating speed is always above the synchronous
speed and the rotor supplies a part of the generated power.
The second mode is called ‘‘synchronous’’ mode, where the
operating speed is the synchronous speed and the rotor does
not supply or receive any AC power but receives a DC power
to act as an exciter. The third is the ‘‘sub-synchronous’’ mode, equations [29]:
where the operating speed is always below the synchronous dψds
speed and the rotor receives a part of the generated power. Vds = Rs Ids − ωs ψqs + (5)
dt
dψqs
Vqs = Rs Iqs + ωs ψds + (6)
dt
dψdr
Vdr = Rr Idr − ωslip ψqr + (7)
dt
dψqr
Vqr = Rr Idr + ωslip ψdr + (8)
dt
ψds = Ls Ids + Lm Idr (9)
ψqs = Ls Iqs + Lm Iqr (10)
ψdr = Lr Idr + Lm Ids (11)
ψqr = Lr Iqr + Lm Iqr (12)
FIGURE 5. The power characteristic of the used WT.
Ls = Lis + Lm (13)
Lr = Lir + Lm (14)
The power characteristic curve between the rotational
speed and the output power for any WT has a MPP that where:
changes according to the wind speed. Therefore, the MPP Vds & Vqs : The d- q axis stator voltages, respectively;
tracking curve specifies the reference power for the control Vdr & Vqr : The d-q axis rotor voltages, respectively;
system of the driven electrical generator according to the ψds & ψqs : The d-q axis stator flux linkages, respectively;
actual speed of the WT [29], Fig. 5. The figure also shows ψdr & ψqr : The d-q axis rotor flux linkages, respectively;
that the proposed turbine is designed to produce always the Ids & Iqs : The d-q axis stator currents, respectively;
maximum power at speeds less than the DFIG synchronous Idr & Idr : The d-q axis rotor currents, respectively;
speed even with different wind speeds. The sub-synchronous Rs & Rr : The stator and rotor resistances, respectively;
operation mode of the DFIG is the part assigned to our study Lm : The magnetizing inductance;
where the wind speed is chosen to be ranged from 13.2 to Ls : The stator leakage inductance;
6 m/s. Table 4 summarizes the main technical specifications Lr : The rotor leakage inductance;
of the proposed wind energy system. Lis : The stator self-inductance;
In this work, the stator voltage oriented control (SVOC), Lir : The rotor self-inductance,
Fig. 6, is used to represent the mathematical model of ωs : The synchronously rotating reference frame angular
the DFIG throughout sets of equations to infer the stator velocity;
active and reactive powers in terms of rotor d-q axis cur- ωr : The rotor angular frequency;
rents as displayed in the following voltage and flux sets of ωslip = ωs − ωr : The slip angular frequency.

192092 VOLUME 8, 2020

H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

FIGURE 6. d-q axis mathematical model of DFIG.

Neglecting the stator resistance Rs [30], the stator active C. THE DG

and reactive powers can be represented in terms of rotor The stochastic and intermittent characteristics of wind and
current as follows: solar powers necessitate the use of another power source
3 Lm that acts as a backup power source to cover the anticipated
Ps = − Vds Idr (15)
2 Ls imbalance in load demand and guarantee the stability and
reliability of the grid. DG system is considered one of the
 
3 Vds Lm
Qs = Vds − − Iqr (16) best selections to be integrated in a MG.
2 ωs Ls Ls
Although it is criticized for the expensive running cost
By rearranging (15) and (16), the following equations can
of fossil fuel and the great environmental damage caused
be deducted:
by generating green gases such as carbon dioxide, it is still
2Ls
Idr = − Ps (17) marked by its large capacity, continuous power supply, strong
3Vds Lm environmental adaptability and flexible operating character-
2Ls Vds istics [31]. The DG system consists of two main parts. These
Iqr = − Qs − (18)
3Vds Lm ωs Lm are; a mechanical part comprising a diesel engine and a
Owing to equations 17 and 18, the DFIG stator active governor system that controls the output frequency and active
power can be controlled by the tuning of the d-axis rotor power. The second part is electrical, consisting of a syn-
current component whereas the q-axis current component chronous generator and an automatic voltage regulator (AVR)
controls the stator reactive power and the terminal voltage. responsible for the terminal voltage regulation and the output
In the proposed system, the DFIG RSC and GSC are man- reactive power.
dated to carry out the control objectives for DFIG. Whereas, Fig. 8 shows the schematic diagram of the considered syn-
the RSC is responsible for regulating the stator terminal volt- chronous generator. It is brushless type containing auxiliary
age to nominal level and prompting the DFIG to extract the synchronous called exciter mechanically coupled to the end
maximum power from the available wind speed depending of the main generator shaft. The main generator excitation
on tuning d-q axis components of the rotor current. The GSC coil is fed by a DC field current resulting from rectification
is taking charge of regulating the DC link voltage and thus it of the AC output voltage of auxiliary exciter by diodes bridge
would manage the power flow at GSC of DC link depending mounted on the rotor. Adjusting the auxiliary exciter field
on the grid side d-q axis current components. Fig. 7 displays current through the AVR, the generator output voltage is
the SVO control scheme for both of GSC and RSC. regulated [32].

VOLUME 8, 2020 192093

 H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

FIGURE 7. Stator voltage oriented (SVO) control schemes for DFIG GSC and RSC.

Table 5 displays the main technical specifications of the simulated system structure. It contains three main controllers:
brushless synchronous generator used in this work. The WT controller, PV controller and ANFIS controller. The WT
selected rated output is asymptotic to the residential load so controller is meant to regulate both the DFIG stator and
that the DG is able to supply the entire load in emergency the GSC terminal voltages to always maintain 575 V as
situations. the main MG bus voltage. It also tunes the DC-link voltage
and extracts the maximum power from the available wind.
III. RESULTS AND ANALYSIS It uses signals from the DFIG stator voltage, stator current,
The proposed system is modeled and simulated using the rotor current, rotor angle, rotor speed, and DC-link voltage as
MATLAB Simulink software package. Fig. 9 displays the inputs, whereas the outputs are the pulses to RSC and GSC.

192094 VOLUME 8, 2020

H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

DG to minimum level. The inputs and output of the ANFIS

controller are explained before.
The proposed system performance with the ANFIS is ini-
tially evaluated under different climatic conditions simulating
the daily natural weather fluctuations of solar irradiation,
ambient temperature and wind speed. The solar irradiation
reaches maximum in the morning then decreases until it fades
away in night. Similarly, the ambient temperature rises to
maximum during midday and falls to minimum during night.
On the contrary, the wind speed is relatively low in daytime
FIGURE 8. Typical architecture of brushless synchronous generators. and increases gradually as the night falls. Moreover, the
simulation has not overlooked the abrupt weather changes in
TABLE 5. The main technical specifications of the brushless synchronous which the wind speed may rise and the sky becomes gloomy,
generators. causing the irradiation and temperature to decrease as shown
in patterns from period t=12 to 14 s.
To ensure the robustness of the proposed system, the step
function is deliberately used in the changes of the three stud-
ied weather conditions. As displayed in Fig. 10(a), the wind
speed is initially set at 12 m/s and starting from t=2 s,
it decreases at a constant rate every two seconds until t=12 s.
Then, there is a sudden increase from 7 to 10 m/s that stays
constant for two seconds then decreases again to 7 m/s at
t=14 s. After that, it increases again at a constant rate every
two seconds until returning to 12 m/s at t=24 s.
The solar irradiation and ambient temperature are set
The PV controller function is to extract the maximum power initially at 0 W/m2 and 15 c◦ , respectively. They are
from the available irradiation. The inputs are voltage and increased at a constant rate then decreased suddenly at t=12 s
current signals from PV terminals while the output is the from 1000 W/m2 and 40 c◦ to 600 W/m2 and 30 c◦ for
pulses to boost converter gate. Lastly, the ANFIS controller two seconds then re-increased to 1000 W/m2 and 40 c◦ at
is responsible for decreasing the fuel consumption of the t=14 s. They are decreased again at a constant rate until

FIGURE 9. Simulink System Structure.

VOLUME 8, 2020 192095

 H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

FIGURE 10. MG Response during variable weather conditions: (a) wind speed, (b) solar irradiation, (c) ambient temperature,
(d) grid Voltage PU, (e) DC link voltage PU, (F) DFIG rotational speed, and (g) DFIG torque.

they are finally returned to 0 W/m2 and 15 c◦ , as shown illustrated in Fig. 11. The electrical and mechanical powers
in Fig. 10(b) and 10(c). of the WT are displayed in Fig. 11(a) and 11(b), respectively.
The first objective of our PMS is to achieve the MG voltage Fig. 11(c) shows the total power generated from the PV which
stability. Fig. 10(d) shows the stator output voltage where it completely or partially feeds the DC link according to the
is remarked that the voltage is not affected by any changes in irradiation availability and the ambient temperature in day
wind speed, irradiation or ambient temperature. The error is time. The DFIG stator output power is illustrated in Fig. 11(d)
always kept within ± 5% of its rated voltage throughout the including both of the power generated from the wind and the
simulation period after system startup. In addition, the DC power drawn from the DC link throughout the RSC shown
link bus voltage is within ± 2% error of its rated voltage as in Fig. 11(e) irrespective of whether it comes from the PV or
displayed in Fig. 10(e). Fig. 10(f) and Fig. 10(g) show the the GSC grid that acts as a bidirectional converter. The GSC
rotational speed and torque of the DFIG. As depicted from supplies the DC bus partially or entirely in the event of lack
the figures, there is a relatively DFIG smooth running without or absence of the PV generated power. Whilst in the case of
overshooting even with different wind speed changes, except increased PV generated power, often due to high irradiation
for the system startup time. availability and high DFIG speeds, the excess power in the
The power sharing between the different power sources DC link drives the GSC to supply the grid instead, as shown
in the proposed MG to supply the 2 MW residential load is in Fig. 11(f).

192096 VOLUME 8, 2020

H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

FIGURE 11. Power sharing of the different power sources in the proposed system: (a) wind turbine electrical power, (b) wind
turbine mechanical power, (c) PV power, (d) DFIG stator power, (e) RSC power, (f) GSC power, (g) electrical DG power, (h)
mechanical DG power, and (i) load power.

On the other hand, Fig. 11(g) displays the electrical power the required 2 MW power with error of no more than 2%
produced by the DG that resulted from the mechanical out- achieving the desired power balance in the MG as the second
put power predicted by the ANFIS controller as shown objective of the work.
in Fig. 11(h). Finally, Fig. 11(i) shows the total power The MPP tracking of both the WT and PV is mandatory to
supplying the residential load. It is proven from the fig- reduce the consumption of the conventional fuel through min-
ure that the proposed system has always been able to provide imizing the DG contribution to the system’s power balance.

VOLUME 8, 2020 192097

 H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

This covers the paper’s third and fourth objectives that are
almost achieved as displayed in Fig. 12 (a) and (b).

FIGURE 12. MPP tracking capability: (a) for the DFIG, (b) for the PV.

FIGURE 14. Comparison of the system performance during variable

FIGURE 13. System Performance under different electrical faults at load weather conditions using (GA-ANFIS), ANFIS and (PSO-ANFIS): (a) MG
terminal: (a) PU voltage stability, (b) the root mean square of the three voltage PU, (b) Load Power, (c) DFIG speed, (d) PV MPP, (e) DFIG MPP.
phases fault current.

The proposed hybrid system and its ability to support stable voltage, takes 80 m.s. whereas the single phase to
the required loads should be checked, even in the event of ground fault achieves stability after only 30 m.s. Fig. 13(b)
any fault. Fig. 13 displays the system performance during shows the root mean square value of the three phases fault
exposure to different electrical faults at load terminals. The current at load terminal during the mentioned short circuit
aforementioned faults include symmetrical and asymmetrical faults. The maximum short circuit current reaches 4460 A
faults which occur and last for 50 m.s. Three phases to after three phases to ground fault and the minimum short
ground, three phases, single phase to ground and two phases circuit current reaches 3250 A after a single phase to ground
faults occur at t = 5, 9, 13, 17 s., respectively. fault.
Fig. 13(a) demonstrates the performance of the proposed Two different optimization algorithms are implemented
system in restoring voltage stability at load terminal and as previously hinted. The results of (GA-ANFIS) and
returns to its nominal value after all electrical faults. The three (PSO-ANFIS) are compared to those of ANFIS. Fig. 14 dis-
phases to ground fault, being the last to reach the accepted plays the comparative assessment of system performance

192098 VOLUME 8, 2020

H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

the voltage stability after all the electrical faults but also it
returned to the nominal value after only 70 m.s. with the three
phases to ground fault and 25 m.s. with the single phase to
ground fault, instead of 80 m.s. and 30 m.s. using the ANFIS.
Whereas the (GA-ANFIS) has failed to restore the nominal
voltage for the three phases to ground and two phases faults.
Moreover, it has also failed to achieve satisfactory results for
the other two faults compared to the ANFIS. The values of
maximum and minimum three phases fault currents at load
terminals for (GA-ANFIS), (PSO-ANFIS) are relatively the
same as ANFIS as depicted in Fig. 15 (b), (c) and (d),

IV. CONCLUSION
A smart PMS based on ANFIS has been proposed for a MG
constructed from hybrid renewable energy system composed
of DFIG driven by WT and PV in conjunctions with DG. The
GA and PSO optimization techniques have been employed
in order to enhance the ANFIS parameters to achieve dif-
ferent objectives with different climatic and fault conditions.
The results demonstrate that the proposed strategy is robust
and could successfully achieve the voltage stability, balance
the generated power and loads and track the MPP of the
RERs. The superiority of the (PSO-ANFIS) is validated com-
pared with the (GA-ANFIS) and the ANFIS. The MG with
the proposed control strategy could successfully manage to
achieve the required goals. The injected power from the RERs
and DG is optimally shared by both the GSC and RSC in
the proposed approach accommodating the uncertainty of the
RERs.

REFERENCES
[1] A. Askarzadeh, ‘‘Electrical power generation by an optimised autonomous
PV/wind/tidal/battery system,’’ IET Renew. Power Gener., vol. 11, no. 1,
pp. 152–164, Jan. 2017.
[2] T. Ma and M. S. Javed, ‘‘Integrated sizing of hybrid PV-wind-battery sys-
FIGURE 15. Comparison of the system performance using (GA-ANFIS), tem for remote island considering the saturation of each renewable energy
ANFIS and (PSO-ANFIS) under different electrical faults at load terminal: resource,’’ Energy Convers. Manage., vol. 182, pp. 178–190, Feb. 2019.
(a) PU voltage stability, (b) & (c) and (d) the root mean square of the fault [3] M. F. Zia, E. Elbouchikhi, and M. Benbouzid, ‘‘Microgrids energy man-
currents drawn in the three phases. agement systems: A critical review on methods, solutions, and prospects,’’
Appl. Energy, vol. 222, pp. 1033–1055, Jul. 2018.
in terms of load power, MG voltage, DFIG speed and [4] L. K. Gan, J. K. H. Shek, and M. A. Mueller, ‘‘Hybrid wind–photovoltaic–
diesel–battery system sizing tool development using empirical approach,
MPP for both PV and WT. These are the five main life-cycle cost and performance analysis: A case study in scotland,’’ Energy
criteria that can fully represent the efficiency of the sys- Convers. Manage., vol. 106, pp. 479–494, Dec. 2015.
tem. In Figs. 14 (a) and (b), the MG voltage and load [5] H. Lin, K. Sun, Z.-H. Tan, C. Liu, J. M. Guerrero, and J. C. Vasquez,
‘‘Adaptive protection combined with machine learning for microgrids,’’
power stability with (PSO-ANFIS) are superior to that of IET Gener., Transmiss. Distrib., vol. 13, no. 6, pp. 770–779, Mar. 2019.
ANFIS whose results are better than (GA-ANFIS). Both [6] Y. E. G. Vera, R. Dufo-López, and J. L. Bernal-Agustín, ‘‘Energy manage-
the (PSO-ANFIS) and (HLA-ANFIS) techniques provided ment in microgrids with renewable energy sources: A literature review,’’
Appl. Sci., vol. 9, no. 18, pp. 1–28, Sep. 2019.
smoother performance of the DFIG than that of (GA-ANFIS) [7] R. Bhaduri, G. R. Saravana, and C. Vaskar, ‘‘Supervisory controller for
as shown in Fig. 14 (c). In Figs. 14 (d) and (e), the MPP power management of microgrid using hybrid technique,’’ Trans. Electr.
of PV and WT are almost similar for the three tech- Electron. Mater., vol. 21, no. 1, pp. 30–47, Feb. 2020.
[8] S. Kewat, B. Singh, and I. Hussain, ‘‘Power management in PV-battery-
niques. However, the curve of (GA-ANFIS) has sometimes hydro based standalone microgrid,’’ IET Renew. Power Gener., vol. 12,
exhibited over-shootings being higher or lower than that of no. 4, pp. 391–398, Mar. 2018.
(PSO-ANFIS) and ANFIS. [9] J. Hong, J. Yin, Y. Liu, J. Peng, and H. Jiang, ‘‘Energy management and
control strategy of photovoltaic/battery hybrid distributed power genera-
The system is further exposed to the same previ- tion systems with an integrated three-port power converter,’’ IEEE Access,
ously described electrical faults. The simulation results of vol. 7, pp. 82838–82847, 2019.
(GA-ANFIS) and (PSO-ANFIS) are compared with those of [10] P. S. Kumar, R. P. S. Chandrasena, V. Ramu, G. N. Srinivas, and
K. V. S. M. Babu, ‘‘Energy management system for small scale hybrid
ANFIS as shown in Fig 15. From the figure, it is concluded wind solar battery based microgrid,’’ IEEE Access, vol. 8, pp. 8336–8345,
that the (PSO-ANFIS) has not only succeeded to restore 2020.

VOLUME 8, 2020 192099

 H. M. Fekry et al.: PMS Based on ANFIS for AC Microgrid

[11] M. K. Senapati, C. Pradhan, S. R. Samantaray, and P. K. Nayak, ‘‘Improved [32] M. Salah, K. Bacha, A. Chaari, and M. E. H. Benbouzid, ‘‘Brushless three-
power management control strategy for renewable energy-based DC phase synchronous generator under rotating diode failure conditions,’’
micro-grid with energy storage integration,’’ IET Gener., Transmiss. Dis- IEEE Trans. Energy Convers., vol. 29, no. 3, pp. 594–601, Sep. 2014.
trib., vol. 13, no. 6, pp. 838–849, Mar. 2019.
[12] S. Koohi-Kamali, N. A. Rahim, and H. Mokhlis, ‘‘Smart power manage- HESHAM M. FEKRY received the B.S. and M.S
ment algorithm in microgrid consisting of photovoltaic, diesel, and battery degrees in electrical power and machines engineer-
storage plants considering variations in sunlight, temperature, and load,’’ ing from Cairo University, Giza, Egypt, in 2006,
Energy Convers. Manage., vol. 84, pp. 562–582, Aug. 2014. and 2012, respectively. He currently works as the
[13] H. Razmi and H. Doagou-Mojarrad, ‘‘Comparative assessment of two Electrical Section Head of Egyptian Propylene and
different modes multi-objective optimal power management of micro-grid: Poly Propylene Company, which is considered one
Grid-connected and stand-alone,’’ IET Renew. Power Gener., vol. 13, no. 6, of biggest petrochemical plants in Middle East.
pp. 802–815, Apr. 2019. His research interests include power system oper-
[14] F. Valencia, J. Collado, D. Saez, and L. G. Marin, ‘‘Robust energy manage- ation and management, renewable energy, gen-
ment system for a microgrid based on a fuzzy prediction interval model,’’ eration and utilization electrical power, artificial
IEEE Trans. Smart Grid, vol. 7, no. 3, pp. 1486–1494, May 2016.
intelligent in energy systems, power quality, micro-grid stability and control,
[15] Y. K. Semero, J. Zhang, and D. Zheng, ‘‘EMD–PSO–ANFIS-based hybrid
smart grids, distributed generation, and AI applications in energy systems.
approach for short-term load forecasting in microgrids,’’ IET Gener.,
Transmiss. Distrib., vol. 14, no. 3, pp. 470–475, Feb. 2020. AZZA AHMED ELDESOUKY received the B.S. and the M.S degrees in
[16] K. Basaran, N. S. Cetin, and S. Borekci, ‘‘Energy management for on- electrical engineering from Suez Canal University, Port-Said, Egypt, in 1989,
grid and off-grid wind/PV and battery hybrid systems,’’ IET Renew. Power and 1995, respectively, and the Ph.D. degree from Bath University, U.K.,
Gener., vol. 11, no. 5, pp. 642–649, Apr. 2017. in 2002. She is currently working as a Professor with the Department of
[17] E. B. Ssekulima, M. B. Anwar, A. A. Hinai, and M. S. E. Moursi, ‘‘Smart Electrical Engineering, Port Said University. Her research interests include
energy management system for optimal microgrid economic operation,’’
power system operation and management, power quality, microgrid stability
IET Renew. Power Gener., vol. 10, no. 7, pp. 885–898, 2016.
and control, smart grids, distributed generation, and AI applications in energy
[18] N. T. Mbungu, R. Naidoo, R. C. Bansal, and M. Bipath, ‘‘Optimisation
systems.
of grid connected hybrid photovoltaic–wind–battery system using model
predictive control design,’’ IET Renew. Power Gener., vol. 11, no. 14, AHMED M. KASSEM received the B.Sc. degree
pp. 1760–1768, Dec. 2017. in electrical engineering from Assiut University,
[19] S. Singh, M. Singh, and S. C. Kaushik, ‘‘Optimal power scheduling of Egypt, in 1991, the M.Sc. degree in electrical
renewable energy systems in microgrids using distributed energy stor- engineering from Eindhoven Technical University,
age system,’’ IET Renew. Power Gener., vol. 10, no. 9, pp. 1328–1339,
The Netherlands, in 1999, and the Ph.D. degree
Oct. 2016.
in electrical engineering from Minia University,
[20] I. M. Alarifi, H. M. Nguyen, A. N. Bakhtiyari, and A. Asadi, ‘‘Feasibility
of ANFIS-PSO and ANFIS-GA models in predicting thermo-physical
Minia, Egypt, in 2006. He is currently a Professor
properties of Al2 O3 -MWCNT/oil hybrid nanofluid,’’ Materials, vol. 12, of electrical power engineering with the Electrical
no. 21, pp. 1–13, 2019. Department, Faculty of Engineering, Sohag Uni-
[21] B. E. Elnaghi and S. A. Selim, ‘‘Performance of double fed induction gen- versity, Egypt, where he has been the Dean of the
erator based wind turbine using adaptive neuro-fuzzy inference system,’’ Faculty of Engineering, since 2014. His research interests include smart
in Proc. IEEE Conf. Russian Young Researchers Electr. Electron. Eng. energy grids, renewable energy management and control, energy storage,
(EIConRus), St. Petersburg, Russia, Feb. 2017, pp. 808–813. power electronics, power systems operation and control, renewable energy,
[22] N. Mahmud, A. Zahedi, and A. Mahmud, ‘‘A cooperative operation of electric drives and their control, electric vehicles, and storage energy devel-
novel PV inverter control scheme and storage energy management system opment and intelligent control applications. He is an editorial board member
based on ANFIS for voltage regulation of grid-tied PV system,’’ IEEE for many journals and a consultant for many industrial companies and a
Trans. Ind. Informat., vol. 13, no. 5, pp. 2657–2668, Oct. 2017. peer reviewer in many transactions, periodicals, and conferences. He has
[23] S. Leonori, A. Rizzi, M. Paschero, and F. M. F. Mascioli, ‘‘Microgrid supervised many Ph.D. and M.Sc. thesis and has published more than
energy management by ANFIS supported by an ESN based prediction 70 research articles. He has got the Sohag University Superiority Award
algorithm,’’ in Proc. Int. Joint Conf. Neural Netw. (IJCNN), Rio de Janeiro, in 2018 in Engineering Sciences.
Brazil, Jul. 2018, pp. 1–8.
[24] A. M. Kassem, ‘‘Modeling, analysis and neural MPPT control design of ALMOATAZ Y. ABDELAZIZ (Senior Member,
a PV generator powered DC motor-pump system,’’ WSEAS Trans. Syst., IEEE) received the B.Sc. and M.Sc. degrees in
vol. 10, no. 12, pp. 399–412, Dec. 2011. electrical engineering from Ain Shams University,
[25] G. Mamatha, ‘‘Perturb and observe MPPT algorithm implementation Cairo, Egypt, in 1985 and 1990, respectively, and
for PV applications,’’ Int. J. Comput. Sci. Inf. Technol., vol. 6, no. 2, the Ph.D. degree in electrical engineering accord-
pp. 1884–1887, 2015. ing to the channel system between Ain Shams
[26] R. Nair and G. Narayanan, ‘‘Emulation of wind turbine system using vector University, Egypt, and Brunel University, U.K.,
controlled induction motor drive,’’ IEEE Trans. Ind. Appl., vol. 56, no. 4, in 1996. He has been a Professor of electrical
pp. 4124–4133, Aug. 2020. power engineering with Ain Shams University,
[27] A. M. S. Yunus, A. Abu-Siada, M. A. S. Masoum, M. F. El-Naggar, since 2007. He is currently the Vice Dean of Fac-
and J. X. Jin, ‘‘Enhancement of DFIG LVRT capability during extreme ulty of Engineering and Technology, Future University in Egypt, Cairo,
short-wind gust events using SMES technology,’’ IEEE Access, vol. 8,
Egypt. He has authored or coauthored more than 400 refereed journal and
pp. 47264–47271, 2020.
conference papers, 25 book chapters, and three edited books with Elsevier
[28] M. I. Mosaad, A. Alenany, and A. Abu-Siada, ‘‘Enhancing the performance
and Springer. His research interests include the applications of artificial
of wind energy conversion systems using unified power flow controller,’’
IET Gener., Transmiss. Distrib., vol. 14, no. 10, pp. 1922–1929, May 2020. intelligence, evolutionary and heuristic optimization techniques to power
[29] B. Subudhi and P. S. Ogeti, ‘‘Optimal preview stator voltage-oriented system planning, operation, and control. He is a member of IET and the
control of DFIG WECS,’’ IET Gener., Transmiss. Distrib., vol. 12, no. 4, Egyptian Sub-Committees of IEC and CIGRE’. He has been awarded many
pp. 1004–1013, Feb. 2018. prizes for distinct researches and for international publishing from Ain
[30] B. Wu, Y. Lang, N. Zargari, and S. Kouro, Power Conversion and Control Shams University and Future University in Egypt. He is the Chairman of
of Wind Energy Systems. Hoboken, NJ, USA: Wiley, 2011. the IEEE Education Society chapter in Egypt. He is a Senior Editor of Ain
[31] B. Zhu, H. Tazvinga, and X. Xia, ‘‘Switched model predictive control Shams Engineering Journal, an Editor of Electric Power Components and
for energy dispatching of a photovoltaic-diesel-battery hybrid power sys- Systems Journal, an editorial board member, an editor, an associate editor,
tem,’’ IEEE Trans. Control Syst. Technol., vol. 23, no. 3, pp. 1229–1236, and an editorial advisory board member for many international journals.
May 2015.

192100 VOLUME 8, 2020