Professional Documents
Culture Documents
Electrical Engineering
(IREE)
Contents
Phase Drift Phenomenon in Dual Active Bridge Converter – Analysis and Compensation 1
by Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
Replacing Copper with New Carbon Nanomaterials in Electrical Machine Windings 12
by J. Pyrhönen, J. Montonen, P. Lindh, J. J. Vauterin, Marcin J. Otto
Computation of Electromagnetic Losses in Double-Rotor Vernier PM Motors 22
with Three Topologies Using TS-FEM
by Huijuan Liu, Yue Hao, Shuangxia Niu, Weinong Fu
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved
(continued)
Photovoltaic System Equipped with a Reliable and Efficient Regulator/MPPT 131
and Energy Flow Controller
by K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Modeling and Control of a Hybrid Microgrid by Multi-Agent System 145
by Lahoussine Elmahni, Lahoussine Bouhouch, Rachid Alaoui, Ali Moudden
Load Frequency Control of Multi Area SSSC and CES Based System Under Deregulation 154
Using Particle Swarm Optimization
by P. Marimuthu, B. Basavaraja, Subhransu Sekhar Dash
Power Line Inspections and Dedicated Flight Envelopes 163
by Sandra C. R. Antunes, K. Bousson, João G. Mota
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – In this paper the mechanisms causing phase drift in the dual active bridge topology
are presented and two compensation methods are proposed to overcome their effects. The phase
drift causes discrepancy between the predicted phase shift and the actual phase shift measured
from the transformer primary and secondary voltages. The error in the phase shift causes the input
power to be larger or smaller than predicted and it can also interfere with the operation of several
modulation methods. By using the phase drift compensation methods presented in this paper the
converter input power can be predicted accurately by using the classical power equation. The
effects of dead-time on the phase drift are also discussed and a method to calculate a suitable
dead-time with respect to the converter operating point is proposed. The value of the phase drift
compensation is presented along with a variable frequency modulation method, which enables
zero voltage switching at a wide load range with a minimal circulating current. Copyright © 2015
Praise Worthy Prize S.r.l. - All rights reserved.
Keywords: DC-DC Power Converters, Dual Active Bridge, Error Correction, Modeling,
Modulation, Power MOSFET, Switched-Mode Power Supply
Nomenclature t Time
tdead Dead-time between the switching leg
Co(tr) Time related output capacitance transistors
Coss Transistor output capacitance tdelay Voltage rise time after a switching instant
DAB Dual Active Bridge tdrift Time difference between the primary and
DSP Digital Signal Processor secondary switching delays
FB Full-bridge V1 Voltage across the leakage inductor,
fsw Switching frequency primary side
HB Half-bridge V2 Output voltage referred to the transformer
hpri Switching configuration parameter, primary
primary side VDS Drain-source voltage
hsec Switching configuration parameter, VFM Variable Frequency Modulation
secondary side Vin Input voltage
ID Drain current Vout Output voltage
Iin Input current ZCS Zero Current Switching
Ipeak Peak current ZVS Zero Voltage Switching
Isw Current at transistor switching instant Phase shift as a percentage of the
Izvs Minimum current required for zero switching period
voltage switching ctrl Phase shift from the feedback controller
Llk Leakage inductance The difference between the actual and
drift
MOSFET Metal-Oxide-Semiconductor Field-Effect desired phase shift as a percentage of the
Transistor switching period
n Transformer turns ratio Phase shift in radians
rad
Npri Transformer primary turns Sum of the feedback controller and
sum
Nsec Transformer secondary turns feedforward phase shifts
P Power
Pctrl Power given by a PI-controller controlling
the transferred power I. Introduction
Pref Power reference
Qtot Total charge in the switching leg transistor Dual Active Bridge converter (DAB) is a galvanically
output capacitances isolated DC/DC converter topology, which has gained a
SiC Silicon Carbide lot of research interest during the last decades due to its
1
Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
2
Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
1
Izvs = 2 A
0 Ipeak = 11 A
Vpri Vpri
Vsec V
-1 sec
Ilk I
lk
-2
0.5 1 1.5 2 2.5 3 3.5
Time [µs]
-1
Izvs = 6.5 A
-2
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time [µs]
0
Izvs = 2 A
-1
Figs. 2. Idealized waveforms for the DAB converter in two cases:
a) V1<V2 and b) V1>V2. The power flow direction is from V1 to V2. Ipeak = 11 A
-2
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
If the delays are exactly the same on the primary and Time [µs]
secondary side, they will effectively cancel each other
Fig. 3. Measured phase drifts for three cases: V1<V2 (top), V1 ≈ V2
and there is no phase drift. However, Fig. 3 illustrates (middle) and V1>V2 (bottom). The input current Iin = 6 A in all cases.
that this may not be the case in practice. The references for primary and secondary voltages are marked with
In Fig. 3 (top) the primary switching leg transistor dashed lines. The converter is using a FB-HB configuration and the
output capacitances are charged and discharged by converter is in zero-voltage switching in all cases. When the delays are
known, the phase drift can be calculated from (7)
current Izvs and the secondary transistor capacitances by
current Ipeak. Since Ipeak is considerably larger than Izvs,
the transistor drain-source voltages can change much The difference between the primary and secondary
faster on the secondary side than on the primary side. delays is not as large as in the case where V1<V2, since
This will cause the effective phase shift to be shorter the primary side Coss charges are larger than the
than intended. In Fig. 3 (middle) the converter is secondary side charges with the voltages used in the
operated near a 1:1 point, where V1 ≈ V2. In this case the measurement.
leakage inductance current is square shaped and both the According to our measurements the dead-time does
primary and secondary sides are switched with same not have an appreciable impact to the phase drift
current. Therefore, the overall phase drift is ideally zero, providing that the dead-time is not too short or too long.
since the switching delays between the primary and Fig. 4 illustrates the effects of a dead-time selection in a
secondary bridges are identical. The identical switching case where Vin = 100 V, Vout = 250 V and Iin = 4 A.
currents do not ensure identical delays in cases where If the dead-time is too short and the transistor is
there are differences between the charges in the primary switched on during the drain-source voltage resonance
and secondary transistor output capacitances. period, the output capacitances are charged and
When the input voltage is higher than the reflected discharged abruptly and this is seen as an overshoot in
output voltage, the primary side transistors are switching the drain current. The effect on the phase drift is that the
at Ipeak and the secondary side transistors at Izvs as transformer voltage transitions become faster and thus
illustrated in Figs. 2 and Fig. 3. The primary side is the delays become smaller. The drawback of a too short
switching faster than the secondary side and the actual dead-time is degraded efficiency, since the transistors are
phase shift is now longer than intended. partially hard switched.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
3
Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
C gs
C gd
Rg
Fig. 5. Zoom of the drain-source voltage (red dashed line) and drain
current (blue solid line) behavior of transistor PA- with 200 ns dead-
time. Since the drain voltage is not changing during t3-t4, there is no
distinct Miller plateau in the gate-source voltage
Qtot
Fig. 4. Measured effects of dead-time on the voltage and current t1 t3 (4)
behavior of transistor PA-. The instant where the gate-source voltage I lk t t1
has risen to the conduction threshold level is marked with Vgs(th)
The current Ilk is not constant during t1-t3 or during the
The optimal instant to turn on the transistor is when delay periods in Fig. 3, but is changing nonlinearly.
the transistor output capacitance is discharged and the However, in most cases it is sufficient to calculate (4)
body diode (or an external freewheeling diode) starts with Ilk(t = t1) being Izvs or Ipeak depending on the
conducting. Since the MOSFET channel is typically a operating mode. In cases where the stored energy in the
less lossy route for the current than the body diode, the leakage inductance is very close to the minimum energy
conduction losses can be minimized by allowing the required for zero-voltage switching, a better
current to commutate to the channel as soon as possible. approximation can be obtained by using the average
If the dead-time is so long that the transistor drain leakage inductance current value during time period t3 in
current changes direction from negative to positive Figs. 2. The length of time period t3 in Figs. 2
before the transistor channel is opened (Fig. 4, bottom) corresponds to the time interval t3-t4 in Fig. 5 and the
the transistor output capacitance begins to charge and the upper limit for its length can be approximated from:
drain-source voltage increases.
If the dead-time is long enough, the transistor voltage I sw Llk
will reach its off state value. This voltage overshoot can t3 t4 (5)
be seen as a voltage sag in the complementary transistor V1 V2
drain-source voltage in the same switching leg. The large
oscillations in the drain-source voltage are likely to In order to ensure zero-voltage switching, the
degrade the efficiency more than the partial hard- transistor PA- gate-source voltage should not be applied
switching during the short dead-time operation. In Fig. 4 before instant t2. After the gate-source voltage has been
the efficiency was 93.4% in the 0 ns case, 93.7% in the applied, the gate-source voltage begins to rise with a time
200 ns case and 92% in the 700 ns case. constant τ, which depends on the gate resistor and on the
The dead-time effects such as voltage sag and gate-drain and gate-source capacitances. If the gate-
overshoot were briefly presented also in [12], but no source voltage rises to the threshold voltage between
guidelines have been given on how to select a proper time instants t3-t4, the transistor is able to conduct the
dead-time with respect to the operating conditions. positive drain current after t4 and there will be no voltage
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
4
Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
overshoots and voltage sags. The dead-time which causes If operating in case where V1<V2 the primary side is
the smallest interference to the current commutation can hard switched while the secondary side switches at peak
now be expressed as: current. In case where V1>V2 the primary side switches at
peak current and the secondary side is hard switched. In
Qtot VDS I sw Llk both cases the delay differences between the primary and
tdead (6) secondary voltages are much smaller than in the soft-
I lk t t1 V1 V2
switching mode (Fig. 3) and thus the phase drift is also
smaller.
where Isw is the current at transistor switching instant
When operating in hard-switching the phase drift may
(either Izvs or Ipeak).The inequality (6) applies only to ZVS
not be an issue at all providing that we are not interested
operation. In practice the leakage inductance current
in the leakage inductance current waveform, but rather
behaves nonlinearly during the switching process and
on the power to be transferred. If the converter is current
cannot change with the rate described by (5) until the
(or power) controlled, the controller will adjust the phase
resonant process during t1-t3 has completed and the
shift so that the current (power) reference is met
voltage across the leakage inductance has reached its
regardless of the phase drift.
maximum value. Therefore, a better approximation for
the maximum dead-time length can be calculated as a
sum of (4) and (5). If the converter operates under hard II.3. Phase Drift Differences Between Different
switching it is enough to ensure that the dead-time is Input-Output Configurations
adequately long to avoid switching leg cross-conduction
and short enough to avoid zero-sequences in the As was mentioned in the introduction, the phase drift
transformer current. phenomenon is emphasized when the primary and
secondary sides have a different switching configuration.
Fig. 7 presents a simulated comparison with a FB-HB
II.2. Phase Drift When Operating in and FB-FB configuration, when using PSpice and
Hard Switched Mode STWA88N65M5 MOSFETs modelled with nonlinear
capacitances.
Fig. 6 provides support for the statement that the
phase drift is smaller in hard switched mode than in zero-
PA+
voltage switching mode.
10
Leakge inductance current [A]
7.5 FB-HB
FB-HB IZVS too FB-FB
5 low
2.5
0
-2.5
-5
-7.5
-10
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time [µs]
8
6 FB-HB hard
Drain current [A]
switching
4
0
FB-FB soft switching
-2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time [µs]
150
Drain-source voltage [V]
100
50
-50
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time [µs]
Fig. 6. Measured phase drifts for two cases: V1<V2 (top) and V1>V2
(bottom). The input current Iin = 4 A in all cases. The references for Fig. 7. An example of the phase drift difference between a FB-FB and a
primary and secondary voltages are marked with dashed lines. The FB-HB converter. The turn-off of transistor PA+ initiates zero-voltage
converter is using a FB-HB configuration and the converter is in hard- turn-on process for the transistor PA-. The leakage inductance current
switching in both cases should be 2.5 A at PA+ turn-off
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
5
Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
The operating parameters in the simulation are: Vin = function of the applied voltage. Fig. 8 illustrates the
100 V, Vout = 250 V, Iin = 4 A, Izvs = 2.5 A. The switching charges obtained for a STWA88N65M5 MOSFET when
frequency that should result in the desired Izvs is 96154 using a PSpice model and the Coss values from the
Hz and the corresponding phase shift is 0.1. The datasheet. For the sake of comparison the charges are
transformer turns ratio in the FB-HB is 1:1, while it is calculated also for a C2M0080120D SiC MOSFET. If a
1:2 in the FB-FB. phase drift calculation is performed for the SiC transistor
In the FB-HB configuration the current that is using the parameters of Fig. 3 (top), the total phase drift
charging and discharging the secondary side transistor is only 0.0017, which is 92% smaller than the phase drift
output capacitances is twice as large as in the FB-FB with the STWA88N65M5.
configuration. Because of this the FB-HB secondary Besides having lower absolute Coss charge values the
switches faster and there is a phase drift like in Fig. 3 SiC charge behavior as a function of drain-source voltage
(top). Due to the phase drift the current Izvs is smaller is much more linear compared to the Si transistor, which
than the specified 2.5 A at the turn-off instant of PA+ and, also decreases the amount of phase drift in various
as a consequence, transistor PA- is hard switched. In the operating points [16].
FB-FB configuration the primary and secondary It is not necessary to measure the charges by using
switching times are closer to each other and the phase only one switching leg. By measuring the delays from a
drift is small. Therefore, the desired turn-off current for complete converter setup, the additional parasitic
PA- is close to the specified value and soft-switching capacitances such the transformer winding capacitances
conditions are met. The phase drift affects the input can be included in the measurement.
current as well. In the FB-HB configuration the input The compensation algorithm can be implemented on a
RMS current was 80% of the reference current, while in DSP by creating a look-up table from the charge and
the FB-FB configuration it was 95%. voltage values, calculating the phase drift from (7) and
adding the phase drift to the phase shift given by the
modulation algorithm.
III. Phase Drift Compensation
In this section two different methods that can be used
to compensate the phase drift effects are introduced. The
phase drift compensation is important when using
modulation methods such as the ones described in [9],
[13] and [14] where the current values at certain time
intervals need to be known.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
6
Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
The charge values can also be used when calculating a those that would be required to achieve ZVS under the
suitable dead-time from (6). A suitable average charge operating conditions. The converter error dynamics are
value could also be used instead of a look-up table similar to the converter overall dynamics and therefore
depending on the output capacitance nonlinearity and the the stability criteria of the compensation controller can be
maximum allowable error in the switching delay derived similarly as would be done for the converter
estimation. power controller.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
7
Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
The component values for the converter are given in shifts are larger than intended. In the 1:1 case the delays
Table I. Fig. 11 presents the measured phase shifts at are nearly identical on the primary and secondary sides
various input and reflected output voltage ratios, when no and therefore the measured phase shifts are close to the
compensation is used. reference. In order to evaluate the impact of phase drift to
the accuracy of the power Eq. (2), the actual phase shift
TABLE I and the input power are measured at various phase shift
COMPONENT VALUES FOR THE DAB PROTOTYPE
references and Vin/Vout values. The efficiency is also
Symbol Quantity Value
Sw Power MOSFETs 6 × STWA88N65M5
measured at each point in order to rule out its effect on
TF Transformer ETD54 N87 the power prediction.
n Turns ratio 1 The curve "calculated" corresponds to the power
n1 Primary winding 25 turns 1050 × AWG44 litz values calculated from (2) using the phase shift reference
n2 Secondary winding 25 turns 420 × AWG46 lits
L1k Leakage inductance 26.4 μH (internal)
and the curve "calculated with losses" is obtained from
Cin Input capacitor 28 × 1μF/450V ceramic cap. (2) using the measured phase shift and by adding the
Cdc1 Voltage doubler capacitor 14 × 1μF/450V ceramic cap. measured power losses to the result.
Cdc2 Voltage doubler capacitor 14 × 1μF/450V ceramic cap.
Cb1 DC-blocking capacitor 10 × 10μF/25V ceramic cap.
500 100
La1 Input choke 63μH, KoolMu 77930/125μ
450 90
Lb1 Input filter damping choke 31μH, KoolMu 77930/125μ
Rb1 Input filter damping resistor 2.2Ω 400 80
La2 Output choke 63μH, KoolMu 77930/125μ 350 70
Efficiency [%]
Lb2 Output filter damping choke 31μH, KoolMu 77930/125μ 300 60
Rb2 Output filter damping resistor 2.2Ω 250 50
200 40
0.12 Measured
150 30
Reference Calculated
0.9:1 100 20
0.1 Calculated with losses
1:1 50 10
Efficiency
1.1:1
0 0
0.08 0 0.02 0.04 0.06 0.08 0.1 0.12
Phase shift
Effective phase shift
650 100
0.06
600 90
550
500 80
0.04
450 70
Input power [W]
Efficiency [%]
400 60
0.02 350
50
300
250 40
0
200 30
150
20
-0.02 100
0 0.02 0.04 0.06 0.08 0.1 0.12 10
50
Phase shift reference
0 0
0 0.02 0.04 0.06 0.08 0.1 0.12
Phase shift
Fig. 11. Measured phase shifts with respect to the reference phase shifts
650 100
at various Vin/nVout ratios. The output voltage Vout is 160 V and the input
600 90
voltages are 74.7 V (0.9:1), 83 V (1:1) and 91.4 V (1.1:1). 550
80
500
450 70
The effects described in conjunction with Fig. 3 can
Input power [W]
Efficiency [%]
400 60
also been seen in Fig. 11. With the voltage ratio of 0.9:1 350
50
the primary side delays are longer than the secondary 300
250 40
delays, which results in smaller phase shifts than 200 30
intended. The reason for the large phase shift error in the 150
20
range of 0.03-0.08 is the change in the converter 100
50 10
operating mode. 0 0
0 0.02 0.04 0.06 0.08 0.1 0.12
Below the phase shift of 0.03 the converter is Phase shift
operating under hard-switching and therefore the
transistor output capacitances are charged and discharged Fig. 12. Measured input powers at various phase shifts with different
abruptly. After the 0.03 mark the converter shifts into input voltage to reflected output voltage ratios. Please see Fig. 11 for
soft-switching, but since the leakage inductance current the voltage levels
is small the zero-voltage transition takes time and the
delays are long (we are allowing the resonance process to It can be seen that by removing the phase drift from
take place by using a long enough dead-time). At larger the calculation and by taking the power losses in to
phase shifts the leakage inductance current is larger and account the calculated power correlates well with the
therefore the phase drift calculated from (7) is smaller. measured power. There should not be large discrepancies
When the voltage ratio is 1.1:1 the secondary delays between the predicted power and the measured power
are larger than the primary delays and therefore the phase like there was in [7], if the converter is operating under
the conditions listed below:
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
8
Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
• The primary and secondary transistors should be able Again, the phase drift compensation is needed to
to produce similar switching times in order to enable switching at zero current. Fig. 13 illustrates the
minimize the phase drift. The factors affecting the operation of the VFM algorithm with and without the
switching times are the transistor output capacitance phase drift compensation. The converter operating point
(the lower the better), the voltage difference between is Vin = 125 V, Vout = 300 V, Iin = 4 A and Izvs = 2.5 A.
primary and secondary sides (the output capacitance
is voltage dependent) and the switching currents.
• The efficiency should be as high as possible to
minimize the effect of power losses.
• The dead-time should be long enough to avoid hard-
switching and short enough to avoid the voltage
overshoot and voltage sag phenomena (Fig. 4).
• The converter has proper input and output filtering so
that we measure the active current instead of the
circulating reactive current [17], [18].
The value of the drift compensation is best
demonstrated by investigating the leakage inductance
current waveforms. If we know that the measured phase
drift is equal to the phase shift given by the power
equation, we also know the lengths of the time intervals
t1, t2 and t3 and the values of Izvs and Ipeak in Figs. 2. The
compensation was used along with a variable frequency
modulation method (VFM), which can be derived from
the idealized DAB model as follows [19]:
1
I ref sign I ref
4 2 I 2 2 I 2
ref ref (8)
h priV2
f sw 1 2
I ref Llk
where:
1 V1 V2
V1 (9) Fig. 13. Measured waveforms of VDS(PA-), inverted ID(PA-), Ilk and Vsec
V V1 V2 with the compensation controller disabled (top) and enabled (bottom).
2 Without the phase drift compensation the value of Izvs is approximately
2 A instead of the desired 2.5 A, which causes the transistor PA- to go
into hard-switching. The transistor current was measured with a PEM
V1 CWT Ultra mini with an output of 50 mV/A
V1 V2
V2 (10)
1 V1 V2 Although the difference in Izvs in Fig. 13 was only 0.5
A, it was enough to make a difference between ZVS
(efficiency 93.44%) and partial hard-switching
and: (efficiency 93.09%). The importance of the drift
sign I ref I zvs h pri (11) compensation increases together with the amount of
phase drift.
In (8) Iref is the reference current for the controller.
The VFM modulation method is explained in more detail
in [19].
V. Conclusion
The purpose of the modulation method is to produce In this paper it was shown that the phase drift
the desired switching current Izvs by varying the phenomenon results from the different switching delays
switching frequency and phase shift accordingly. In order between the primary and secondary bridges. When using
to ensure that the algorithm produces the desired Izvs and transistors with small output capacitances (such as wide
zero-voltage switching, the actual phase shift measured band-gap MOSFETs) and if the differences between the
from the transformer primary and secondary voltages primary and secondary switching currents are small, the
must match the phase shift given by the algorithm. phase drift can be negligible. The phase drift is of
The algorithm can also be used to obtain zero current importance if such modulation methods are used, where
switching by setting Izvs = 0. the leakage inductance current at a certain time instant
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
9
Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
10
Vesa Väisänen, Jani Hiltunen, Raimo Juntunen
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
11
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – Electrical machines have significant improvement potential. Nevertheless, the field is
characterized by incremental innovations. Admittedly, steady improvement has been achieved, but
no breakthrough development. Radical development in the field would require the introduction of
new elements, such that may change the whole electrical machine industry system. Recent
technological advancements in nanomaterials have opened up new horizons for the macroscopic
application of carbon nanotube (CNT) fibres. With values of 100 MS/m measured on individual
CNTs, CNT fibre materials hold promise for conductivities far beyond those of metals. Highly
conductive, lightweight and strong CNT yarn is finally within reach; it could replace copper as a
potentially better winding material. Although not yet providing low resistivity, the newest CNT
yarn offers attractive perspectives for accelerated efficiency improvement of electrical machines.
In this article, the potential for using new CNT materials to replace copper in machine windings is
introduced. It does so, firstly, by describing the environment for a change that could revolutionize
the industry and, secondly, by presenting the breakthrough results of a prototype construction. In
the test motor, which is to our knowledge the first in its kind, the presently most electrically
conductive carbon nanotube yarn replaces usual copper in the windings. Copyright © 2015 Praise
Worthy Prize S.r.l. - All rights reserved.
Keywords: Electrical Machine, Winding Material, Carbon Nanotube Yarn, Machine Design,
Efficiency Improvement, Motor Prototype Construction
12
J. Pyrhönen, J. Montonen, P. Lindh, J. J. Vauterin, Marcin J. Otto
MS/m and its resistivity temperature coefficient is [15]. Because of the “one-dimensional” and symmetric
3.886×10−3/K [12]. For silver, the respective values are structure of CNT fibre, the charge carriers can travel
63MS/m and 3.8×10−3/K [11]. But the improvement along the nanotubes almost without “scattering”, which
found by using silver instead of copper would be minor is a phenomenon that is commonly referred to as
because the density of silver is higher than that of copper, “ballistic transportation”. The almost absence of
as a result of which the machine mass might even scattering helps CNTs to carry very high current
increase. The respective value for aluminium are 35 densities, theoretically in the order of J = 100 MA/cm2
MS/m and 3.9×10−3 /K. Aluminium has thus a lower [16]. CNTs could then offer, again in theory, a thousand
conductivity, [10] but aluminium is widely used e.g. in fold increase in current density compared to that of
the rotors of squirrel cage induction motors [11]. Normal metals such as copper. Conductivity values in the range
construction steel has a conductivity of about 4 MS/m of 100 MS/m have been measured on individual single-
and electrical steel qualities can have conductivity values walled armchair CNTs at room temperature, resistivity
in the range of 1.5 – 2 MS/m [13]. values in the range of 1.0×10-8m [11], and the
Despite the fairly high conductivity of copper, a major resistivity temperature coefficient is negative; –0.2×10−3
share of the machine losses still occur in the copper /K [17].
windings; this phenomenon is denoted as copper losses So far, the electrical properties of CNT
[13]. A typical industrial motor has a minimum rated macrostructures such as yarns and sheets are
point efficiency of 91.4 %. At the rated operating point it considerably lower than those of individual CNTs [18].
converts about 1040 W into heat losses. Of these losses, However, from a theoretical viewpoint and assuming that
ca. 45 % are created in the stator winding, 27 % in the the intrinsic high conductivity nature holds from the
rotor windings, 21 % are iron losses and the rest are microscopic level of CNTs to the macroscopic level of
friction and additional losses [13]. In an induction motor, CNT yarn, it would seem that CNT yarn does, in
the copper losses are responsible for ca. 72 % of all the principle, not have a pre-defined upper limit for
losses in the machine, which corresponds to 750 W of conductivity. This means that the electrical conductivity
heat power. properties of CNT yarn are still improving day by day as
Considering that the demand for valuable metals such a result of continuous research and development efforts
as copper is likely to increase tenfold as developing and advances in the field. It is therefore possible to
economies surge ahead and that the production of copper assume that the conductivity of CNT yarn will soon catch
requires energy-intensive extraction from its ore, copper up with and even exceed the conductivity of copper, and
is increasingly considered an energy-expensive and thereby open up enormous development potential for the
carbon-intensive material for use in rotating electrical electrical machine industry.
machinery. Moreover, the levels of cradle-to-gate energy Recently, the Japanese-Dutch company Teijin Aramid
consumption in the production of copper may be very developed a new spinning technology in collaboration
high. Copper has also a relatively high metal density with Rice University in the USA [19]. CNTs are spun to
(8960 kg/m3) which means that copper products may be form multi-fibre yarn by wet spinning technology. The
heavier than comparable other products. industrial applications of the new CNT yarn are,
Increasing the amount of copper in a machine does however, still in their infancy. The problem of spinning
hence not necessarily improve the overall energy highly conductive multifibre yarn is with the purity of the
efficiency of the machine but may negatively affect the CNTs. The Teijin Aramid wet spinning process itself
carbon footprint of the machine. There also remain the results in high-quality joints between individual CNTs
risks associated with growing import dependency on and, as such, has potential to produce high conductivity
metals. The EU, for instance, is dependent on imports to fibre as soon as the problem of CNT purity is solved
meet 50% of its copper demand [14] and, hence, also rigorously. Though the wet spinning technology is yet at
dependent on globally set high copper prices. Aiming at laboratory scale, the techniques provide promising
upgrading the performance and efficiencies of electrical outlooks, and are beginning to bridge the gap between
machines, higher-conductivity wires are searched for basic research and development for CNT-fibre yarn to
windings. meet industrial needs. This assumption is based on
observed similarities between development processes of
CNT yarn and Twaron Aramid yarns, the latter of which
III. Emerging Alternatives from has been successful in scaling up the technology to
Nanomaterials Research industrial volumes. Today, it seems that finally it is
Where metals seem to have hit a ceiling, the new within reach to improve the conductivity in the electrical
nanomaterials may offer a case for more powerful machine windings by replacing metals with new CNT
improvement of electrical machines. Theoretical fibre materials.
physicists and experimentalists have worked on carbon Finally, the most recently achieved resistivity for CNT
nanotube (CNT) materials for over two decades, but only yarn would be low enough to allow for testing yarns in
recently the use of carbon nanotubes (CNTs) in concrete different electrical apparatuses. Applications of carbon
applications and in constructions has become acceptable. conductors have been reported previously, but these
Armchair CNTs are, by structure, highly conductive experiments predominantly focus on the use of CNT
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
13
J. Pyrhönen, J. Montonen, P. Lindh, J. J. Vauterin, Marcin J. Otto
materials in electric devices. Examples of these are the operating temperatures than what we are used to apply in
supply of current to an incandescent lamp [19], a 400 Hz electrical machine design today. This is due to the fact
small transformer with CNT windings [20]. that there will be no increase of Joule losses as the
Nanostructured carbon materials for electrical temperature increases. As far as we use steel in the
applications have also been tested in energy storages magnetic circuit, its losses get smaller as the temperature
[10], [21] and in solar cells [12]. The results showed that increases because the eddy currents will be limited by the
nanomaterials could have significant potential as increased lamination resistivity. The eddy current losses
electrode materials e.g. as a cathode for wearable and are inversely proportional to the resistivity of the
lightweight energy storage devices [22]-[24]. laminations [9]. The higher operating temperatures will
naturally result in new challenges regarding the
insulation and permanent magnet materials. Present-day
IV. Exploiting CNT Conductivity: An insulation materials seldom can operate in temperatures
Important Step in the Development higher than 220 C [11] and NdFeB-based permanent
of Electrical Machines magnets should normally be operated below 150 C.
Let us assume that in the near future researchers will
If it will be possible to achieve higher conductivity
succeed in manufacturing 100 MS/m CNT yarn
values by substituting copper for new CNT fibre material
conductors. Copper conductivity is 59.6 MS/m at room
in the windings of an electrical machine, most probably
temperature with a resistivity temperature factor of
the efficiency of the machine will be improved or the
3.886×10−3/K; at 150 C it is only 39.6 MS/m. If eddy
dimensions and mass of the machine will be decreased,
currents and circulating currents decrease the practical
or both. Such development, consequently, will bring
conductivity further by 20 % we stay at the level of about
about the need for change in the machine properties
32 MS/m while CNT yarn could maintain its 100 MS/m
determination and design. It will involve fundamental
change in the overall design principles of electrical at 150 C. This means that, in practice, we could reduce
machines. the Joule losses of a future CNT-yarn electrical machine
In present-day electrical machines the operating by two thirds if the machine geometry is kept the same.
Already today, the properties of available CNT yarn do
winding temperature is normally in the range of 120 C.
allow important niche development, especially in the
Under load the copper DC-conductivity decreases to the
domain of turbomachinery for instance for special high-
level of 42.9 MS/m. This is 72 % of the copper
frequency high-temperature compressor applications.
conductivity at room temperature. In CNT yarn
Furthermore, CNT fibre materials also allow
conductors the resistivity should stay about constant
manufacturing of lightweight and strong electrical
when their temperature has risen to 120 C. Hence, CNT
conductors. The density (CNT 1500 kg/m3) of CNT-
yarn windings may have multiple operating-temperature
fibre yarn is about one sixth the density of copper.
conductivity values compared with the conventional
Already now, the specific conductivity of CNT-fibre
copper windings.
A further benefit of CNT yarn conductors is that they yarn (/ [Sm2/kg]) is close to the level of the specific
consist of very thin sub-conductors which should conductivity of copper. It is therefore realistic to expect
considerably limit the possibility of skin effect at that in the near term ongoing efforts to leverage the
electrical machine frequencies. The yarn inherently properties of CNTs will accelerate the development of
resembles litz wires. Moreover, if we may assume that CNT yarn such that will be able to conduct electricity
future commercial CNT yarn will be manufactured from with the same mass as copper conductors do nowadays.
multifibres which can be easily transposed, no circulating Though the significance of such development will be
currents should appear either. Both skin effect and minor in electrical machines where the space reserved for
circulating currents can be very harmful in traditional conductors is limited, nevertheless, it might greatly
high-current windings and lead into increased AC- benefit the development of other, special applications.
resistance. In the copper conductors of a high-power And last but not least, the availability of carbonic
electrical machine there is always the problem of materials is of major importance for the further
possible skin effect and circulating currents which make development of resource-efficient CNT winding
the AC-resistance of copper conductors significantly machines. Carbon materials are among the most
higher than the DC-resistance value. It is not uncommon abundant materials on the Earth. As such, the presently
that the resistance factor (kR = RAC/RDC) gets values of high price of carbon-based conductors should decrease
150 % in higher-frequency applications [11]. In high- with the growth of industrial applications and
frequency, high-temperature applications we might, commercial use. At present, the price of copper is in the
therefore, theoretically reach a situation where the AC- range of 10 €/kg and its density is about six times the
conductance of a carbon conductor is even three-times density of CNT yarn conductor material [25]. It is not
the conductance of a similar size copper conductor. This expected that CNT yarn conductors will soon beat copper
should bring about a remarkable change in electrical conductors. However, we may expect that in the short
machine design. term CNT materials will enable the creation and
It should then also be possible to design carbon- development of new applications and new devices, there
winding electrical machines that can operate at higher where CNT wires can, in niche applications, bring
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
14
J. Pyrhönen, J. Montonen, P. Lindh, J. J. Vauterin, Marcin J. Otto
advantage over copper in light weight, flexibility, However, in our motor application a much lower
bending fatigue resistance, resistance to corrosion, high current must be used because the present-day insulation
strength and high modulus of elasticity. materials should not be exposed to too high operating
Yet today, since the manufacturing development of temperatures. The 26 AWG yarn type was selected for
carbonic conductors is still at laboratory scale, in the test motor. This type has a cross-sectional surface of
practice, the conductivity of CNT yarn conductors is still 0.1280 mm2, an equivalent single wire diameter of 0.405
at a modest level; that is 3.4 MS/m, which is in the range mm and corresponds to CNT yarn of approximately 2000
of construction steels’ conductivities. dtex.
However, poorer conductive steel can be and is often Each motor conductor is made of ten parallel 2000
used in solid-rotor induction motors as rotor conductor dtex yarns. The motor coil conductors are 1.2 m long,
material. The presently poorer conductivity of available and their measured resistance is 0.4 , yielding only 2.4
CNT yarn should therefore not be a barrier for looking at MS/m average conductivity whilst the predicted
the use of CNT yarn for special electrical machine conductivity was 3.4 MS/m. The yarn has also a small
applications. positive temperature coefficient (0.0008/K) whilst we
With this article, we suggest that is now time to focus expected the yarn to have zero or a small negative
research efforts on the largely untapped potential for temperature coefficient. The conductivity level of the
advance in electrical machine technology offered by sample conductors is thus very low. Therefore, the yarn
already available and future CNT materials as these hold can be used only for demonstration purposes; that is to
the promise of significantly improving the conductivity show the potential of CNT yarn for application in an
in electrical machines. In the following section we electrical machine.
present the results of an experiment we have recently Regarding the test motor manufacturing, we faced the
conducted. We have designed and constructed the – to problem that Teijin Aramid CNT-yarn is not provided
our knowledge – first electrical machine in which we with insulation. We solved the problem by preparing a
have used the presently most electrically conductive tape of parallel conductors (with insulating Twaron yarns
CNT yarn to replace the conventional copper wires in the as side protection) on insulating aramid (Twaron) paper
windings. strip. The turn-to-turn insulation is thus made by using
The overall aim of the project was to launch a this kind of insulating paper. We recognized that, while
comprehensive study on the feasibility of improving having parallel conductors on a flat tape, this might
rotating electrical machine efficiencies and performance increase the risk of circulating currents in the windings.
by replacing copper with highly conductive new CNT However, as the winding material has high resistance
materials. One of the main objectives of the experiment we decided to ignore this problem for the test machine.
was to demonstrate that there is an urgent need for Fig. 1 illustrates the conductor material on an Aramid
collaboration between the fields of CNT materials paper tape. For transport purposes the wires are wound
research and engineering research. Collaboration needs on a large paper cylinder. The final conductor is spread
to reach a level where scientists and manufacturers will on a 9 mm wide Aramid paper strip so that the surfaces
work together so as to increase better understanding of of the flat conductor are insulated just on one side.
the benefits and challenges regarding the development The cross-sectional surface dimensions of the flat
and application of carbon nanomaterials to improve the conductor with single-sided insulation are 0.5 9 mm2.
energy and resource efficiency of rotating electrical
machines.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
15
J. Pyrhönen, J. Montonen, P. Lindh, J. J. Vauterin, Marcin J. Otto
TABLE I
CNT-YARN PERMANENT MAGNET SYNCHRONOUS MACHINE
DESIGN PARAMETERS IN GENERATING
Parameter Value Fig. 3. Single stator lamination sheet on the left and the assembled PM
STATOR MAGNETIC CIRCUIT rotor on the right. The magnetic circuit form is designed to reserve
Stator stack length lsFe [mm] 42 rectangular spaces for the cross-sections of the aramid-strip-insulated
Stator lamination space factor kFe 0.96 conductors
Stator core material SURA NO10
Stator inner diameter Ds [mm] 25 A three-tooth two-pole machine is the simplest
Stator stack outer electromagnetic diameter Dse possible three-phase machine. It has the benefit that the
75
[mm]
Number of stator slots Qs 3 winding procedure can be simplified when insulated
Number of slots per pole and phase q 0.5 conductors are used, as shown in Fig. 1.
ROTOR The slots are large and therefore the motor could take
Rotor outer diameter Dr [mm] 23 much more conducting material in the stator. The CNT-
PM cylinder diameter, DPM [mm] 20
Rotor PM length, lrPM [mm] 50 yarn space factor is now 15 % and could theoretically be
increased to about 50 %. This should also significantly
PM material (N38UH) remanence Br @ 120 C
1.15 lower the carbon losses. However, the insulation system
[T] illustrated in Fig. 1, does not allow a better space factor
PM relative permeability r at temperature of 120
degrees Celsius
1.05 for the conductors.
PM material coercive field strength Hc [kA/m] 871 To have a high space factor for the CNT yarn, the yarn
Rotor construction: Cylindrical PM is located should be insulated somehow similarly as enamelled
inside a stainless steel tube with 1.5 mm wall copper wires nowadays are.
thickness.
The rotor of the machine has only one cylindrical
STATOR WINDING
Winding type Fractional slot, NdFeB-magnet in a stainless steel tube which allows to
concentrated, non- keep the overall design of the machine as simple as
overlapping, single possible. In a machine with such high resistance in the
layer q = 0.5
winding, PM excitation is, in practice, the only possible
Winding connection: star connected
Number of pole-pairs p 1 option.
Stator coil-turns in series per phase Ns 7 The rotor is made of four parts, two shaft parts, a
In each coil turn there are 10 strands of 0.4 stainless steel tube and a cylindrical magnet with 50 mm
mm diameter. length and 20 mm diameter.
COOLING
Cooling method Air cooling The stator stack end finger plates were manufactured
to have semi-circular cross-sectional areas which makes
the end winding bends as smooth as possible.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
16
J. Pyrhönen, J. Montonen, P. Lindh, J. J. Vauterin, Marcin J. Otto
range of PMech = 1 W.
1
0.5
0.4 0
0 5 10 15 20 25 30
0.3 Harmonic order
Air gap flux density [T]
0.2
0.1 Fig. 5. Induced voltage at no-load rated speed 15000 min-1. In addition
to the fundamental the voltage contains some fifth and seventh
0
harmonic voltages. The fundamental RMS no-load voltage is EPM =
-0.1 3.96 V
-0.2
5
-0.3
-0.4 4
-0.5 3
0.6
0
-1
0.5
-2
Air gap flux density [T]
0.4 -3
-4
0.3
-5
0.345 0.3455 0.346 0.3465 0.347 0.3475 0.348 0.3485 0.349 0.3495 0.35
0.2
Time [s]
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
17
J. Pyrhönen, J. Montonen, P. Lindh, J. J. Vauterin, Marcin J. Otto
TABLE II
3 SUMMARY OF THE MACHINE SIMULATED AS GENERATOR
Parameter Absolute value
2 Rated Voltage, V 3.2/5.5
Rated Current, A 2.3
1 Apparent power, VA 33.8
Rated Input Mechanical Power, W 43.7
Current [A]
0.345 0.3455 0.346 0.3465 0.347 0.3475 0.348 0.3485 0.349 0.3495 0.35
Time [s]
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
18
J. Pyrhönen, J. Montonen, P. Lindh, J. J. Vauterin, Marcin J. Otto
5
5
4
4
3
Phase U
1
1 0
0 -1
-1 -2
-2 -3
-3 -4
-4 -5
0 0 .0 05 0 .0 1 0 .0 15 0 .0 2
-5 Tim e [s]
6
5
4
Induced voltage [V]
4
2
Voltage [V]
3
0
2 -2
1 -4
-6
0
0 5000 10000 15000
Speed [rpm]
-8
0 0.005 0.01 0.015 0.02 0.025 0.03
Fig. 11. Induced voltage RMS value as a function of rotational speed Time [s]
This article is the first in its kind to introduce a break- The article scans the environment and indicates some
through approach to the use of new carbon nanomaterials future perspectives for potential applications of carbon
to enable the development of a new generation of nanotube yarn in electrical rotating machines where
rotating electrical machinery. significant efficiency improvement can be achieved.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
19
J. Pyrhönen, J. Montonen, P. Lindh, J. J. Vauterin, Marcin J. Otto
The topicality of the proposed feature article lies in CNTs. We believe that then a new era in electrical
the description of the potential for integration of new and machine development will start. We may mention that, in
greener carbon nanomaterials into electrical machine the course of writing this article, the best laboratory
development and innovation of industrial production results measured on Teijin Aramid CNT fibre yarn have
lines within the context of the transition towards a more been reported to be now in the range of 10 MS/m. Such
resource-efficient economy. In other words, it shows how an increase of conductivity value represents already a
it may now be within reach to considerably improve the significant improvement compared to the CNT yarn
energy efficiency of those millions of electrical rotating material we used in the test motor.
machines that keep industry and society on the move With the introduction of replacing copper with new
while coupling such development to the important and carbon nanomaterials in electrical machine windings we
topical issues of natural resource savings, carbon savings may have the solution to overcome the challenge of agile
as well as cost savings issues for more sustainable and development in electrical machines. CNT materials may
economic growth. The articles uses the experimental help realize the vision of building greener, better
results from prototype building and testing research the resource efficient, and better performing electrical
objective of which was to design and measure the machines. Notwithstanding the risk that further
world’s first electrical motor applying a textile material; technology development in CNT materials will be less
windings made of carbon nanotube yarn. A prototype of fast as may be expected today, the new carbon
a rotating electrical machine using CNT fibre -based nanomaterials can be of considerable significance for
winding material was designed and tested. important niche development in rotating electrical
It demonstrates that CNT yarn has the potential to machinery, there where CNT materials have advantage
replace copper in the machine windings. over copper in light weight, flexibility, high bending
Furthermore, based on the findings from the prototype fatigue resistance, resistance to corrosion, high strength
construction and associated research investigations, the and high modulus of elasticity.
results presented in this article strongly indicate that by
replacing copper with advanced conductive CNT
materials in the machine windings it may be possible to Acknowledgements
improve the efficiency and performance properties of the The authors wish to acknowledge the support of: Ron
future electrical machine. ter Waarbeek for spinning the CNT fibres; Peter Akker
Copper has continued to be used as an effective and Arjan van Haren for assembling the conducting tape;
conductor in the electrical machine windings for over a Dr. Hanneke Boerstoel for commenting on this paper,
century. However, in the search for higher energy Teijin Ltd. for manufacturing and providing the CNT
efficiency we must look for a solution that will decrease yarn; Antti Suikki and Jouni Ryhänen for arranging the
the losses of electrical machines. Copper losses are the manufacturing of the test machine mechanics; Harri
dominating losses in many types of electrical machines Loisa for constructing the winding; Manu Niukkanen and
and, hence, even a slight improvement in the Jan-Henri Montonen for arranging the control logic of
conductivity of the winding materials may revolutionize the converter; Kyösti Tikkanen for arranging the
electrical machine development and call for redesign of laboratory setup; LUT for funding the prototype design,
complete production lines in the entire industry. manufacturing and testing of the prototype machine.
The experimental results presented here show well
enough that further work is needed to pave to way for
supreme CNT materials to be integrated into production References
lines that will change the electrical machine industry. [1] Commission Regulation (EC) No 640/2009 of 22 July 2009.
Theoretically, the DC-resistivity of CNT yarn Available at
conductors in a future CNT-generation motor is, in eurlex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2009:191:
practical operating temperatures, significantly lower than 0026:0034:EN:PDF
[2] F.W. Geels, “Technological transitions as evolutionary
that of copper conductors. Very thin sub-conductors also reconfiguration processes: a multi-level perspective and a case-
remove all practical skin effect and circulating current study”, Research Policy, vol. 3, 2002, pp. 1257–1274
related problems which further may increase the benefits [3] Z. Wenliang, T.A. Lipo, K. Byung-Il, “Material-Efficient
achieved from using CNT yarn conductors. A significant Permanent-Magnet Shape for Torque Pulsation Minimization in
SPM Motors for Automotive Applications”, IEEE Trans. on
reduction in the conductor resistivity should affect the Industrial Electronics, vol. 61 n. 10, 2014. pp. 5779-5787.
design of machines most significantly. It is then only a [4] A.Boglietti, A.El-Refaie, O.Drubel, A.Omekanda, N.Bianchi,
question of optimization if we shall exploit the properties “Electrical Machine Topologies: Hottest Topics in the Electrical
of CNT yarn material to design smaller and more Machine Research Community”, IEEE Trans. on Industrial
Electronics magazine, vol. 8 n. 2, 2014. pp. 18-30.
lightweight or more energy efficient machines.
[5] A.Tenconi, S.Vaschetto, A.Vigliani, “Electrical Machines for
The CNT-yarn spinning technology is ready for High-Speed Applications: Design Considerations and Tradeoffs”,
manufacturing of high-conductor materials. As soon as IEEE Trans. on Industrial Electronics, vol. 61 n. 6, 2014. pp.
theoretical physicists, materials researchers, engineer 3022-3029.
[6] G.-A. Capolino, A. Cavagnino, “New Trends in Electrical
researchers and experimentalists will closely together
Machines Technology – Part I”, IEEE Trans. on Industrial
they may succeed in selecting the best of pure metallic Electronics, vol. 61 n. 8, 2014. pp. 4281-4285.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
20
J. Pyrhönen, J. Montonen, P. Lindh, J. J. Vauterin, Marcin J. Otto
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
21
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – This paper focuses on the electromagnetic losses computation and comparative
analysis of the proposed three topologies vernier permanent-magnet (VPM) motors by using
time-stepping finite element method (TS-FEM). A dynamic core loss model of numerical analysis
considering two-dimensional (2-D) magnetic properties to grasp the magnetic phenomenon inside
the motors is adopted to compute the core losses of the three VPM motors. The core loss model can
provide reasonable accuracy and all necessary parameters are either directly available or
extracted from manufacturer-provided loss curves. The flux distribution, copper losses, core losses
and efficiency of the three proposed VPM motors have been investigated, and the core loss is the
important loss part of the proposed flux-modulation VPM motors. Copyright © 2015 Praise Worthy
Prize S.r.l. - All rights reserved.
Keywords: Electromagnetic Loss, Core Loss, Efficiency, Vernier Permanent Magnet (VPM)
Motor, Time-Stepping Finite Element Method (TS-FEM)
22
Huijuan Liu, Yue Hao, Shaungxia Niu, Weinong Fu
low-speed operation [3]-[5]. There are many new proposed VPM motors. TS-FEM is employed to analyze
structures of VPM motor have been studied, such as the transient performance of three VPM motors and the
double rotor PM motors, double stator motors and core loss calculation method is employed to compare the
multi-rotor multi-stator PM motors [6]-[7]. core losses of the proposed motors.
For the purpose of reducing the volume and improving The proposed TS-FEM is used to study a
the torque density, three novel topologies VPM motors are two-dimensional problem defined in the plane.
presented and designed. And in order to achieve higher The Maxwell’s equations applied to the air-gaps, iron
efficiency for the proposed VPM motors, one method is to cores and PM regions is the following diffusion equation:
decrease the losses in the motors. But first of all, we must
compute the electromagnetic losses accurately, and there φ m m y
are two components electromagnetic losses in VPM φ 0 x (1)
t y x
motors, one is the copper loss, the other is core loss, and it
is still challenging to predict core loss under transient
conditions in the design of motors and the model of power The left term of Eq. (1) only exists in PM regions. The
loss has been continuously under study for over one magnetic field equation in the region of the stranded
hundred years. Currently, the electric machine design windings can be expressed as:
engineers depend heavily on finite element analysis
Df N f
(FEA), requiring an accurate core loss formula, in order to φ if 0 (2)
achieve better design of efficient machines. S f am p
Usually, machine design engineers select the loss
computation method based on two final goals: accuracy To achieve maximum energy efficiency of the three
and efficiency. Different core loss computation can be VPM motors in design is to minimize the losses including
applied to the FEA in order to achieve the design goals, the core and copper losses. In other hands, the power
such as the micromagnetic eddy current model [8]-[11], losses may be used as an indirect measure of the
the dynamic hysteresis core loss model [12], [13] and the temperature rise of the three motors as a first order of
traditional core loss model [14]. In general, only approximation.
numerical solution of micromagnetic equations based on The Maxwell’s tensor equation is used to calculate the
3-D Landau–Lifshitz–Gilbert equations, together with magnetic force on the rotor. The core losses calculation in
Maxwell’s equations, can provide appropriate and the frequency domain is based on the following equation
accurate physical interpretation of magnetic material [15]:
losses [8]. However, solution of the complicated pcore phys peddy padd (3)
micromagnetic equations requires extensive
computational time, which is prohibitive for engineering
The three types of loss in the time domain can be
design [10]. Accordingly, a dynamic core loss model [15]
computed by:
of numerical analysis considering two-dimensional (2-D)
magnetic properties to grasp the magnetic phenomenon
Lz T db
inside the motors is adopted to compute the core losses in phys 0 hirr dt d dt
the three VPM motors, and the details of this dynamic T
core loss model will be presented in section II. (4)
L T db dby
Compared with traditional VPM motors, the three z 0 hx irr x hy irr d dt
T
dt dt
novel flux-modulation double rotor VPM motors make
more full use of the inner space and save more copper
material depending on their double rotor topology and Lz db 2 db 2
y
fractional slot concentrated winding [16-17]. peddy K eddy x d (5)
According to the PM location, they can be referred as 2 2 dt dt
motor I (PM in rotor), motor II (PM in stator and rotor),
and motor III (PM in stator). They have the similar 3
operation principles. In this paper, the electromagnetic Lz db 2 db 2 4
x y d
core losses, copper losses, efficiency, air-gap radial flux padd
Cadd K add dt dt
(6)
density and flux distributions of three VPM motors are
investigated and compared quantitatively. Three VPM
motors’ performances are analyzed by using The coefficient Keddy and Kadd can be calculated
time-stepping finite element method (TS-FEM), and their according to the core loss curves using linear regression
respective advantages and disadvantages are investigated. algorithm. According to [15], Cadd=8.763363.
In the analysis, the phase resistor of the stator winding
in the effective region (excluding the end winding) is:
II. TS-FEM and Electromagnetic
Losses Analysis 1 L 1 L
r N s Nc N s N c2 (7)
There are two air-gaps and two rotating bodies in the 3 Sc 3 Ss
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
23
Huijuan Liu, Yue Hao, Shaungxia Niu, Weinong Fu
Then the total copper loss is: two sets of 18 pole-pair PMs are added and embedded in
the stator slots, then the PM pole-pairs p1 is 17, stator teeth
L L p2 is 18, and the pole-pair number of the stator windings p
pCu 3I 2 r I 2 N s N c2 I 2 N s2 N c2 (8) is 1. In Fig. 1(c), the rotor of motor III is designed with
Ss Sw
salient structure, and each side of the stator surface is
mounted with 18 pole-pair PMs as motor II, then the PM
Sw is the cross-sectional area of all the slots and it will pole-pair number p1 is 18, and the number of steel
be the same in all the motors being studied for consistent segments p2 is 17 and the pole-pair number of the stator
comparison. NsNc is the total number of conductors in all windings p is 1.
the slots of each motor.
From Eq. (8) it can be seen that for the three VPM
motors, owing to the stator and the stator windings, and
the rated stator current (4A) are all the same, then the
ampere-conductors INsNc are the same and the copper loss
are the same.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
24
Huijuan Liu, Yue Hao, Shaungxia Niu, Weinong Fu
80
60
40
20
0
0 5 10 15 20
Time (ms)
(a) Motor I
(a) Motor I
140
120
CoreLoss (W)
100
80
60
40
20
0
0 5 10 15 20
Time (ms)
(b) Motor II
(b) Motor II
80
CoreLoss (W)
60
40
20
0
0 5 10 15 20
Time (ms)
(c) Motor III
Figs. 2. Core-loss vesus time at different operation of three VPM motors (c) Motor III
100
80 TABLE II
60 PERFORMANCE COMPARISON OF THREE VPM MOTORS
40
Motor type Motor I Motor II Motor III
20
Core loss at locked rotor (W) 9 11 10
0
0 5 10 15 20
Core loss at no-load (W) 75 107 45
Time (ms)
Core loss at full-load (W) 83 109 19
Copper loss (W) 0.05 0.05 0.05
Fig. 3. Core-loss vesus time at full-load operation of three VPM motors Efficiency (%) 91 90 96
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
25
Huijuan Liu, Yue Hao, Shaungxia Niu, Weinong Fu
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
26
Huijuan Liu, Yue Hao, Shaungxia Niu, Weinong Fu
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
27
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – In recent years, permanent magnet machines have been widely used in many
applications such as electric vehicles and hybrid applications. The present work carries out the
design of a surface mounted permanent magnet motor (SMPM) with fractional slots. At a first
stage, an analytical model of the SMPM is developed. At a second stage, based on this derived
model, an optimization procedure of the machine, using the Genetic algorithm, is applied to find
optimal machine geometry in agreement with a set of design criteria. The used objective function
is to maximize the machine efficiency.
A SMPM with 18 slots and 16 magnets is chosen. At a last stage, the obtained analytical results of
the optimal design are validated with finite element analysis (FEA). Based on this validation a
great concordance between analytical and FEA results is shown. Copyright © 2015 Praise
Worthy Prize S.r.l. - All rights reserved.
28
A. Mansouri, H. Msaddek, H. Trabelsi
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
29
A. Mansouri, H. Msaddek, H. Trabelsi
The general machine layout in which are reported the Stator yoke height, hsy:
different parts, is illustrated in Figure 3.
D0 D 2hss
hsy (2)
2
D 2hsw
bss1 bts (3)
Qs
D 2hss
bss 2 bts (4)
Qs
D
s (5)
Qs
Fig. 3. Machine layout
Br kleak
II.4. Machine Sizing Bm (9)
k
Referring to Figs. 1 and 2, we can deduce the 1 r c
lm
following equations:
Inner stator diameter, D:
kleak is determined according to [12], kc is the Carter
factor [14]:
D Drc 2lm 2 (1)
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
30
A. Mansouri, H. Msaddek, H. Trabelsi
s
kopenbss1 Acu f s Asl (18)
kopenbss1 5
fs considers both slot and conductor insulations as well
Let us consider a rectangular shape of the airgap flux as gaps [13]. Typical values of the fill factor are in the
range of 0.4:0.5. The area of the copper conductor is
density, the fundamental airgap flux density B̂ can be given by the following equation:
then determined as follows [13]:
Acu
4 Acond (19)
B̂ Bm sin (11) nc
Once the number of conductors and the rated current
III.2. Analytical Calculation of the Flux Density are known, we evaluate then the copper losses:
in the Teeth and Yoke
Pcu R ph I 2ph (20)
The second flux densities to determine are those of the
stator teeth and yoke created by the permanent magnets.
A special attention should be given to these parameters. The resistance per phase winding can be calculated
Because an overestimation than expected leads to the [8]:
saturation of the stator teeth and yoke and hence a high 120 nc c 2 L Qs
magnetic flux leakage and high iron losses. R ph cu (21)
rcu2 m
The flux densities in the stator teeth Bst and yoke Bsy
are connected to the stator teeth width, bst, and stator
yoke height, hsy, through the following equations [15]: The coil pitch τc is calculated as [8]:
DBm Q D
bst (12) c int s (22)
Bst QS Nm Nm
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
31
A. Mansouri, H. Msaddek, H. Trabelsi
TABLE IV
IV. Optimization Procedure MAGNETIC CONSTRAINTS
Variable Constraint
As described in section II, the machine geometry is
stator teeth flux density Bst 1.5 ≤ Bst ≤ 2
completely defined through 10 parameters. To identify stator yoke flux density Bsy 1 ≤ Bsy ≤ 1.5
these unknown parameters, an optimization procedure rotor yoke flux density Bry 1.3 ≤ Bst ≤ 1.6
must be realized. In the literature several objectives Airgap flux density B̂ 0.8 ≤ Bst ≤ 1.5
function can be focused, minimization of the permanent
magnet material weight, minimization of the total losses
Concerning the used objective function is the
or maximization of the efficiency. In our work, the
maximization of the machine efficiency η defined as
chosen objective function is the maximization of
follows:
machine efficiency. An optimization design procedure is
generally realized throughout three major steps. It begins Pout
(29)
by identifying the unknown machine design variables. Pout Ptot
Secondly the machine constraints and the adjective
function are defined and finally an appropriate solver is Pout is rated output power of the machine and Ptot
applied to find the optimum geometry satisfying all the defines the total losses given by:
requirements.
Ptot Pteeth Pyoke Pcu (30)
IV.1. Machine Design Variables As it has been mentioned, the optimization problem is
The machine design parameters are the input variables the maximization of the efficiency, however the used
of the optimization problem from which the machine solver is the genetic algorithm, then the optimization is
geometry is totally defined. In our work, these transformed to a minimization problem as follows:
parameters are: the airgap length, the magnet height, the
pole angle, the stator exterior diameter, the stator teeth objective function = minimum 1 (31)
width, the stator slot height, the slot opening over the slot
width ratio, the slot wedge height, the active machine The flowchart illustrating the optimization procedure
length and the rotor yoke height. is presented in Fig. 4.
All these parameters with their ranges are reported in
Table II.
V. Results
IV.2. Constraints and Objective Function At first the optimization problem is solved by means
In optimization PM machine design several types of of the genetic problem. All simulations are carried out by
constraints can be used. In our case, two classes of applying the Matlab 2012 software. The obtained design
constraints were used. variables, subject of the optimization, are presented in
Table V.
TABLE II
DESIGN VARIABLE RANGES TABLE V
OPTIMUM DESIGN VARIABLE
Design variable Range
airgap length δ (mm) [1;2] Design variable Range
magnet height hm (mm) [2;4] airgap length δ (mm) 1.2
pole angle α (°) [18;21] magnet height hm (mm) 2
stator exterior diameter D0 (mm) [150;200] pole angle α (°) 20
stator teeth width bts (mm) [6;8] stator exterior diameter D0 (mm) 171
stator slot height hss (mm) [18;25] stator teeth width bts (mm) 6.6
slot opening ratio kopen [0.5;0.6] stator slot height hss (mm) 22
slot wedge height hsw (mm) [1;2] slot opening ratio kopen 0.55
active machine length L (mm) [150;200] slot wedge height hsw (mm) 1.6
active machine length L (mm) 165
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
32
A. Mansouri, H. Msaddek, H. Trabelsi
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
33
A. Mansouri, H. Msaddek, H. Trabelsi
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
34
A. Mansouri, H. Msaddek, H. Trabelsi
Authors’ information
Laboratory of Computer, Electronics & Smart Engineering Systems
Design Engineering School of Sfax, Tunisia.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
35
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – The principal aim of this paper is devoted to real time implementation of sliding mode
control (SMC) strategy based on indirect field oriented control (IFOC) for an induction motor
(IM) drive. The proposed method is a powerful tool to use for a robust nonlinear control law to
model uncertainties and disturbances. Three controllers qualified as sliding mode controllers are
investigated. The design and implementation for a 1.5 kW induction motor drive are presented and
completely carried out using a dSPACE 1104 controller board based on digital signal processor
(DSP). Experimental results show the improvement in performances of the proposed control
algorithm. Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved.
Keywords: Induction Motor Drive, Sliding Mode Control Strategy, dSPACE 1104 Controller
Board
36
A. Essalmi, H. Mahmoudi, A. Bennassar, A. Abbou, Y. Zahraoui, M. Akherraz
III. Sliding Mode Control By derivation of Eq. (10) and we replace the fifth
equation of the system (1) in (10), we obtain:
The sliding mode control is to bring the trajectory
state and to evolve it on the sliding surface with a certain
dynamic to the equilibrium point. To determine the * pLm i TL ƒ
S (11)
rd sq
JLr J J
sliding surface, we take the form of general equation
given by J. J. E. Slotine [4]:
We define the control law:
n-1
d
S x e x (6) isq isqeq isqn (12)
dt
During the sliding mode and in permanent regime,
with e(x) is denote the error of the controlled greatness, λ
is a positive coefficient and n is the order of the system. S S 0 , isqn 0 . The equivalent control
action can be defined as follow:
JLr * TL ƒ
isqeq J J (13)
pLm rd
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
37
A. Essalmi, H. Mahmoudi, A. Bennassar, A. Abbou, Y. Zahraoui, M. Akherraz
A. Sliding Mode Currents Design A 1.5 kW three phase squirrel cage induction
The sliding surfaces for the control of current errors machine, 2 poles, A/3.4A, 220V/400V, 50Hz and
are: 1423rpm;
S isd i*sd isd (15) A static power electronics convertor from Semikron
composed of a diode rectifier and a three leg voltage
source IGBT inverter;
S isq i*sq isq (16) Current are measured using LEM sensors and both of
them are then transformed to be a voltage ranging
By derivation of Eq. (15) and (16) and we replace the from 0 to ±10 volts which will be the input of A/D
first and second equation of the system (1) in (15) and respectively.
(16) respectively, we obtain: The rotor speed is measured by an incremental
encoder, which is mounted at the rotor shaft end. A
K 1 digital filter is used to remove the noise from the
S isd isd
*
isd s isq r vsd (17) speed signal;
Tr Ls
A dSPACE 1104 DSP board with ControlDesk
software used for real time control of the induction
1
S isq isq
*
isq s isd K r r vsq (18) motor drive;
Ls A personal computer with MATLAB/Sirnulink
software to design a simulate induction machine
We define the control law: control.
The motor parameters used in this experimental
vsd vsdeq vsdn (19) investigation are indicated in Appendix. Figs. 2, 3 show
the experimental platform scheme used.
The results presented in the Figures 4-6 below show
vsq vsqeq vsqn (20) experimental results of the sliding mode control strategy
for a squirrel cage induction motor. Here, the reference
During the sliding mode, S isd S isd 0 , flux is kept constant at the nominal value 1 Wb.
* K
vsdeq Ls isd isd s isq rd (21)
Tr
vsqeq Ls isq
*
isq s isd K r rd (22)
vsqn K q sign S isq (24)
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
38
A. Essalmi, H. Mahmoudi, A. Bennassar, A. Abbou, Y. Zahraoui, M. Akherraz
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
39
A. Essalmi, H. Mahmoudi, A. Bennassar, A. Abbou, Y. Zahraoui, M. Akherraz
The SMC strategy show good results in transient [9] A. Essalmi, H. Mahmoudi, A. Bennassar, M. Akherraz, A.
Abbou, neuro-genetic sensorless sliding mode control of a
conditions in speeds. The stator phase current in the
permanent magnet synchronous motor using Luenberger observer,
induction motor remains sinusoidal and takes appropriate Journal of Theoretical and Applied Information Technology, vol.
value. Speed closed-loop control using a sliding mode 53 n. 2, 2013, pp. 236-246.
controller is good for the different cases considered. [10] S. El M.Ardjoun, M. Abid, A.G. Aissaoui, A. Naceri, A robust
fuzzy sliding mode control applied to the double fed induction
machine, International Journal Of Circuits, Systems And Signal
Processing, vol. 5, n. 5, 2011, pp. 315-321, NAUN, USA.
V. Conclusion [11] Y. Sayouti, A. Abbou, M. Akherraz, H. Mahmoudi, Real-time
DSP implementation of DTC neural network-based for induction
This paper presents experimental results of sliding motor drive, PEMD, vol. 3, April 2010, pp. 1-5.
mode control strategy for a 1.5 kW induction motor [12] A. Abbou, T. Nasser, H. Mahmoudi, M. Akherraz, A. Essadki,
drives. The efficacy of these controllers has been verified Induction motor controls and Implementation using dSPACE,
through its hardware and MATLAB/Simulink WSEAS TRANSACTIONS on SYSTEMS and CONTROL, vol. 7, n.
1, 2012, pp. 26-35.
implementations. All experimental results confirm the [13] S. Amamra, L. Barazane, M.S. Boucherit, A. Cherifi, Inverse
good dynamic performances of the developed drive fuzzy model control for a speed control induction motor based
systems and show the validity of the proposed method. dSPACE implementation, Modern Electric Power Systems 2010,
The features of this method, applied to the non linear Wroclaw, Poland, paper P50.
[14] M. Boussak, Implementation and experimental investigation of
systems, that it ensures a fast dynamic response. sensorless speed control of permanent magnet synchronous motor
The proposed controllers are implemented in real time drive, IEEE Transactions on Power Electronics, vol. 20, n. 6,
by employing a dSPACE system with DS1104 controller 2005, pp. 1413–1422.
board based on digital processor Texas Instruments [15] A. Abbou, H. and Mahmoudi, Real time implementation of a
sensorless speed control of induction motor using DTFC strategy,
TMS320F240 DSP. Proceedings of the International Conference on Multimedia
Computing and Systems ICMCS’09, Ouarzazate, Morocco, April
2-4, 2009, pp.327-333.
Appendix [16] Moutchou, M., Abbou, A., Mahmoudi, H., Induction machine
speed and flux control, using vector-sliding mode control, with
TABLE A1 rotor resistance adaptation, (2012) International Review of
INDUCTION MOTOR PARAMETERS Automatic Control (IREACO), 5 (6), pp. 804-814.
Rated power 1.5 kW [17] Khiari, B., Sellami, A., Andoulsi, R., Mami, A., A novel strategy
Voltage 400V Y control of photovoltaic solar pumping system based on sliding
Frequency 50 Hz mode control, (2012) International Review of Automatic Control
Pair pole 2 (IREACO), 5 (2), pp. 118-125.
Rated speed 1440 rpm [18] Houda, B.M., Khadija, D., Said, N.A., New discrete sliding mode
Stator resistance 4.5 Ω controller with predictive sliding function, (2013) International
Rotor resistance 4.54 Ω Review of Automatic Control (IREACO), 6 (4), pp. 529-536.
Inductance stator 0.754 H [19] Abdellah, A., Abdelhafid, A., Mostafa, R., Combining sliding
Inductance rotor 0.7536 H mode and linear quadratic regulator to control the inverted
Mutual inductance 0.7421 H pendulum, (2013) International Review of Automatic Control
Moment of Inertia 0.015 kg m2 (IREACO), 6 (1), pp. 69-76.
Viscous friction 0.003N s/rad [20] Mezouar, A., Terras, T., Fellah, M.K., Hadjeri, S., Composite
sliding mode control of induction motors using singular
perturbation theory, (2012) International Review of Automatic
Control (IREACO), 5 (6), pp. 901-910.
References [21] Guisser, M., El-Jouni, A., Abdelmounim, E.H., Robust sliding
mode MPPT controller based on high gain observer of a
[1] V.I. Utkin, Sliding mode control design principles and
photovoltaic water pumping system, (2014) International Review
applications to electrical drives, IEEE Trans. on Industrial
of Automatic Control (IREACO), 7 (2), pp. 225-232.
Electronics, vol. 40, n. 1, February 1993.
[22] El-Jouni, A., Guisser, M., Salhi, M., A sliding mode optimization
[2] S. Müller, Doubly fed induction generator systems, IEEE Industry
of a photovoltaic pumping system, (2014) International Review
Applications Magazine, vol. 8, n. 3, May-June 2002, pp. 26-33.
on Modelling and Simulations (IREMOS), 7 (3), pp. 466-473.
[3] H. De Battista and al, Sliding mode control of wind energy
[23] Anil Kumar, T., Venkata Ramana, N., Design of sliding mode
systems with DOIG-power efficiency and torsional dynamics
observer based optimal controller for load frequency control in
optimization, IEEE Trans. Power Systems, vol. 15, n. 2, May
multi area deregulated thermal system connected with HVDC
2000, pp. 728-734.
link, (2013) International Review on Modelling and Simulations
[4] J.J.E. Slotine, W. Li, Applied nonlinear control, Prentice Hall,
(IREMOS), 6 (2), pp. 543-553.
USA, 1998.
[24] Benbouzid, M., Beltran, B., Ezzat, M., Breton, S., DFIG driven
[5] P. Lopez, A.S. Nouri, Théorie Elémentaire Et Pratique De La
wind turbine grid fault-tolerance using high-order sliding mode
Commande Par Les Régimes Glissants, Springer, 2006.
control, (2013) International Review on Modelling and
[6] Shaojing Wen and Fengxiang Wang, Sensorless Direct Torque
Simulations (IREMOS), 6 (1), pp. 29-32.
Control of High Speed PMSM Based on Variable Structure
Sliding Mode, Proceedings of Electrical Machines and Systems,
ICEMS-IEEE, International Conference on, 2008, pp. 995 – 998.
[7] F. Blaschke, The principle of field orientation as applied to the Authors’ information
new transvektor closed-loop control system for rotating-field
machines, Siemens Rev., vol. 34, 1972, pp. 217–220.
[8] Bennassar, A., Abbou, A., Akherraz, M., Barara, M., A LEEP, Department of Electrical Engineering, Mohammed V University,
comparative study of IP, fuzzy logic and sliding mode controllers Mohammadia School’s of Engineers, Rabat, Morocco.
in a speed vector control of induction motor, (2013) International E-mails: abbou@emi.ac.ma
Review on Modelling and Simulations (IREMOS), 6 (6), pp. 1865- yassin.zahraoui@gmail.com
1871. akherraz@emi.ac.ma
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
40
A. Essalmi, H. Mahmoudi, A. Bennassar, A. Abbou, Y. Zahraoui, M. Akherraz
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
41
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Supawat Supakwong
Abstract – Traditional array signal processing techniques have been relying on the use of
received signal's second-order statistic for many years. However, it suffers with some fundamental
limitations. Studies of array processing based on higher-order statistic has been proposed aiming
to overcome these limitations. This paper is aimed to assess the array performance enhancement
when using higher-order statistic from the differential geometry perspective. Defined as the locus
of all array response vectors over a set of signal parameters, the array manifold's geometrical
shape and properties are known to be crucially important in characterizing the array
performance. In this paper, the geometry of an array manifold associated with a higher-order
statistic is investigated using of the concept of virtual sensor array. Performance analysis is
presented to examine the array performance enhancement both in terms of the Cramer Rao lower
bound and the array detection and resolution capabilities. Copyright © 2015 Praise Worthy Prize
S.r.l. - All rights reserved.
42
Supawat Supakwong
maximum number of sources is not limited by the formulation is first presented. This includes the array
number of sensor elements N , 2) separate non-Gaussian signal modelling, and the analysis of the received signal's
signals from additive white noise, and 3) provide higher-order statistic. Then, in Section III, the differential
robustness to modelling error. geometry framework is proposed to investigate shape and
Subsequently, a generalization of subspace-based DF properties of manifold associated with higher-order
methods such as 2q MUSIC was proposed [7]. statistics.
The performance analysis will be presented in Section
The essential understanding come from the concept of
IV focusing initially on the choice of circular covariance
virtual-sensor array proposed in [8], where it draws the
matrix arrangement that will optimize the array
relationship between an underlying of array's higher- performance, before examining the array performance
order statistics and the second-order statistics in terms of enhancement in terms of the CRB, detection and
the virtual sensor array to show the virtual increase of the resolution thresholds for various orders of statistics.
array aperture as well as the number of array's virtual Finally, the paper is concluded in Section V.
sensors. The analysis offers important insights into the
mechanism of higher-order methods from the traditional
second-order statistic approaches. II. Problem Formulation
To assess performance of array processing methods, To proceed, the signal modelling is first presented in
either for second-order and higher-order statistics, it is this section. Then, the analysis on the use of higher-order
essential to evaluating through the characteristic/property statistic (HOS) of the received signal is examined.
of an array manifold. By definition, array manifold is the
locus of all array response vectors (manifold vectors)
over the set of signal parameters. For the case of a single II.1. Signal Modelling
parameter, the locus of these manifold vectors represents Consider an array system consisting of N sensor
a curve embedded in the complex space, while the locus
of two-parameter manifold vectors represents a surface elements located at r r x , r y , r z N 3 measured in
[9]. half wavelength, where r x, r y , and r z are column
The fact that the array manifold incorporates all
information regarding the array and signal environments vectors of the sensor position in a Cartesian coordinate
enables the manifold to completely characterize the system with the array centroid is located at the reference
system. Important array performance parameters such as point 0 , 0 , 0 .
the Cramer Rao lower bound (CRB), the array's detection The array is operated in the presence of M narrow-
and the array resolution capabilities (the ability of an band point sources propagated from the array far-field
array to correctly estimate the number of sources, and to with a plane wave propagation is assumed from azimuth-
distinctively obtain signal parameters) can be expressed elevation i ,i ,i 1,...,M with the azimuth
in terms of the manifold's intrinsic parameters.
Issues of array uncertainty analysis and array i 0,180 is measured anti-clockwise from the
ambiguity that is caused by linear dependence amongst positive x axis, and the elevation i 0 ,90 is
source's manifold vectors [10]-[12] as well as the array
measured anti-clockwise from the x y plane.
formation [13] are also related to the shape of array
manifold. Differential geometry has been used as a tool The baseband received signal x t N 1 is
to investigate the geometrical shape and characteristic of expressed as:
the manifold [14].
So far, the majority of the work has focused on the T
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
43
Supawat Supakwong
xx x t x H t Note that the circular covariance matrix 2q,x l in
M (3) Eq. (5) can be arranged in different ways depending on
c2,m a i ,i a H i ,i 2 N
i an arbitrary integer l, 0 l q, which has a direct impact
i 1
on the array performance. The study in [3] showed that
with c2 ,mi denoting the second-order autocumulant of the the number of signals that the array can process is
maximized when the integer l is given by:
message signal mi t , N is an N dimensional
H
identity matrix, and is the Hadamard operation q / 2,
l
if q even;
(7)
(conjugate-transpose). Subspace-based DF algorithms, q 1 / 2, if q odd.
such as MUSIC, exploits the eigenstructure of this data
covariance matrix xx in search for a set of Another advantage from processing signals using
corresponding manifold vectors that lies in the signal higher-order statistic is the robustness to the presence of
subspace (likewise, orthogonal to the noise subspace) Gaussian noise. By comparing the expression of the
with the cost function , given by: 2q order circular covariance matrix in Eq. (5) with the
expression of the second-order covariance matrix in Eq.
, a H , N HN a , (4) (3), one can draw a relationship between the term
a , l a , * q l in Eq. (5) and the vector
i i i i
where N denotes a matrix whose columns contain the
a i ,i in Eq. (3).
associated noise eigenvectors, obtained from the
eigendecomposition of the covariance matrix xx . This indicates an important relationship between the
second-order and a higher-order statistic in a way that the
The use of second-order statistic in array signal
2q order circular cumulant matrix can be thought of as
processing has several limitations. For instance, the
number of sources the methods can process cannot being the second-order covariance matrix of a virtual
exceed N 1 . In addition, the array performance is sensor array with an effective manifold vector ae i ,i
degraded due to the presence of Gaussian noise, the given by:
modelling error, as well as its inability to deal with non-
Gaussian signals that practically employed in digital l * q l
communications. ae i ,i a i ,i a i ,i (8)
hence:
II.2. Higher-Order Statistics of the Received Signal
To overcome the limitations caused by the second- M
H
order statistics, array processing methods based on 2 q,x c2q,m ae i ,i ae i ,i
i
(9)
2q order statistics of the received signal, where q 1, i 1
44
Supawat Supakwong
with the vector k k i ,i for the sake of simplicity eigenvalues corresponding to the signal subspace, and
second, a matrix of eigenvectors with zero eigenvalue
and the operator denotes the Kroneckor addition.
corresponding to the noise subspace. The generalized
Consequently, the Kronecker products over l terms of 2q MUSIC algorithm is based on a search for effective
a i ,i can be simplified to: manifold vectors in Equation (8) that lie in the
corresponding signal subspace, or similarly orthogonal to
a i ,i
l
exp j rT rT rT k the noise subspace.
exp j rT
l
k (11) III. Differential Geometry of an
Array Manifold
T l Array manifold represents a locus of all manifold
where r rT rT rT . Furthermore, the
vectors over the set of signal parameters. It can be
Kronecker product between two conjugated manifold viewed as a geometrical object embedded in a high-
vectors a* i ,i gives: dimensional complex space with the dimensionality is
defined by the total number of sensors in an array. In this
section, differential geometry will be used as a tool to
a* i ,i a* i ,i exp jrT k exp jrT k analyze the shape and characteristic of the manifold.
Understanding the geometrical properties of an array
(12)
exp j rT rT k manifold is crucially important as it is directly related to
the array processing performance in all aspects ranging
from array detection, parameter estimation, to reception.
Hence, the Kronecker products over q l terms of To proceed, we will first review the basic
the conjugated manifold vectors is simplified to: characteristics of array manifold with respect to the
second-order statistics before later extend the framework
q l to analyze the array manifold due to higher-order
a* i ,i
exp j rT rT rT k statistics.
q l
(13)
exp j rT k
III.1. Characterizing Array Manifold Curve
For the simplicity in our analysis, let's assume that an
Therefore, we can derive the expression of ae i ,i N element sensor array under consideration has a
in Eq. (8) to: linear configuration (i.e. r y r z 0 ) and all sources are
impinging from the same elevation, assumed 0 .
ae i ,i exp j r T l
k Under these assumptions, the expression of the array
response vector in Eq. (2) can now be expressed as a
q l single-parameter function of an azimuth direction :
exp j rT k
(14)
a exp j rx cos , N1 (16)
exp j r k
e
T
T l T q l
: 0 ,180
re r r (15)
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
45
Supawat Supakwong
1 da
s s 1 s 1 d (19)
1 d
ds da
s (20)
d d
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
46
Supawat Supakwong
According to the expression of the effective manifold in the similar manner. For instance, the principle
vector corresponding to the higher-order statistics of the curvature of the effective manifold curve can be found
received signal in Eq. (8), the locus of these effective from:
manifold vectors ae forms the corresponding curve
embedded in an NVS dimensional complex space, i.e.: d 2 ae s
i , e s ae '' s (25)
ds 2
Ae ae NVS 1 , : 0 ,180 (22)
Given that ae s exp j r x,VS s r x,VS v from
where NVS represents the total number of virtual sensors [9], thus:
listed in Eq. (15). Several points can be addressed
regarding this effective manifold curve. First, we should d d
draw the fundamental difference between the effective
i,e s
ds ds
exp j rx,VS s rx,VS v
curve and the manifold curve corresponding to the
second order statistic. jrx,VS jrx,VS exp j rx,VS s rx,VS v
This effective manifold curve represents the locus of
r x,VS
2
manifold vectors corresponding to the virtual sensor
array with the position listed in Eq. (15), which is
different from the actual sensor array position. Regardless of the order 2q , the curvature i,e s of
The number of virtual sensors NVS is always greater
the corresponding effective manifold curve Ae remains
than the number of actual sensors N implying that the
constant. This shows that the 2q order effective
effective manifold curve is embedded in the higher-
dimensional complex space. The dimensionality of the manifold curves all have hyperhelical shape embedded in
space depends on the array configuration as well as the an NVS dimensional complex space.
order of statistic 2q. On the other hand, the two curves
share some relationships, which can be related using the
concept of virtual sensor array. That is, to analyze an
IV. Performance Analysis
array manifold curve corresponding to a higher-order The performance enhancement of subspace-based
statistic of the sensor array r x , it is equivalent to methods when using higher-order statistics shall be
studying the second-order manifold curve with respect to investigated in this section.
the virtual sensor array r x,VS . Consequently, the For the purpose of comparison, three performance
indicators will be used. The first is the popular Cramer
presented framework can be used to assess the Rao lower bound (CRB) that expresses a lower bound on
geometrical shape of the effective manifold curve. First, the error covariance matrix of an angular estimate ˆ ,
the effective manifold's arc length can be found from:
which can be expressed in terms of the curve's
geometrical property, in case of a single emitter [9], as:
d ae
se 0 d
d 1
CRB 2
(26)
d
exp j rx,VS cos d 2 SNR L s
0 d
The bound is shown to be inversely proportional to the
0 j rx,VS sin exp j rx,VS cos d (23)
signal-to-noise ratio SNR, the number of snapshots L,
as well as the square of the rate-of-change of manifold
0 rx,VS sin d
curve s . Two other array performance parameters
rx,VS 1 cos are the array detection and array resolution threshold
accordingly. An array detection is referred to the ability
with the rate-of-change of arc length is given by: of an array system to uniquely estimate the presence of
sources in an environment [9], In general, the total
da number of sources can be obtained from the estimated
se r x,VS sin (24) dimensionality of the noise subspace using the property
d of the eigendecomposition associated with the received
signal's second order statistic. However, due to
Furthermore, the system of coordinates u i,e s uncertainties surrounding the channel conditions or
measurements, the dimensionality of the noise subspace
together with the set of curvatures i,e s can be derived
can be difficult to estimate precisely.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
47
Supawat Supakwong
An ultimate's array detection represents a It should be noted, however, that in [8] the choice of l
performance's lower bound of the system's ability to is chosen according to the maximization of the number of
determine the number of sources under a situation when virtual sensors.
two sources are closely located. Typically, the bound is Whereas for our finding here, the choice is based on
given in the form of a product between the signal-to- optimization of the array processing performance. Next,
noise ratio SNR1 and the number of snapshots L, consider an array of 2 sensors separated by half-
expressed as [9]: wavelength, i.e.:
T
2 1 1
1 P rx , (30)
SNR1 L det thres 2
1 1 (27) 2 2
2s P2
measured in half-wavelength. This illustrates a scenario
where P1 and P2 represent the signal powers for the two for a small sensor array employed in a limited space,
sources respectively. Note also that the detection such as in an aircraft or portable devices. Using the
capability is inversely proportional to the square of the second-order statistics, the array can process no more
arc length separation s, which denotes the separation than a single source.
This showed a serious drawback especially in the
of two sources in the form of arc length. multisource environment. However, with a higher-order
On the other hand, the ultimate's array resolution statistic, the 2q subspace-based methods can be used to
capability is referred to a performance's lower bound of
the system to distinctively obtain signal parameters for process a larger number of sources. For the purpose of
signals located close together in space, which can be analysis, let's examine the characteristics of the manifold
expressed in terms of the manifold's differential curve corresponding to different orders of statistic. The
geometry as: 4th , 6th ,8th and 16th orders of statistic are chosen in this
example to compare with the conventional second-order
4 statistic.
2 P1
SNR1 L res thres 1 4 (28) This corresponds to the integer q being 2 ,3, 4 and 8
1 P2
s 4 ˆ 12 accordingly. Using Eq. (15) to obtain the virtual sensor
N array position, the result is shown in Table I together
with the total number of virtual sensors and the virtual
It is seen that, in addition to the arc length separation, array aperture. Observe the increase in the total number
the resolution threshold also depends on the estimated of virtual sensors and the virtual array aperture when
first curvature of the manifold curve ̂1 at s. increasing the order of statistic.
To proceed, let's first examine the issue of the circular Table II summarizes the basic characteristics of each
covariance matrix 2q,x l arrangement and its curve including arc length, rate-of-change of arc length,
the total curve length, and the principle curvature for
performance on the array processing. It was shown in Eq. different order of statistic.
(5) that the matrix 2q,x l can be arranged in a number Note that these curves are embedded in a completely
different dimensional complex space depending on the
of ways depending on an arbitrary l, where 0 l q.
number of virtual sensors for each order.
In this example, let's consider a linear array of three Several points can be addressed regarding the
sensors located at: geometrical property of the corresponding manifold
curves. First, the curvatures of all manifold curves of all
T
r x 0.5,0.1,0.4 (29) orders are constant at all points indicating that the higher-
order manifold curves also have the hyperhelical shape.
This is an important finding which will aid the
measured in half wavelength. Suppose the 8th order analysis for a number of array processing applications
statistic method is employed q 4 . Hence, the integer such as the analysis of manifold ambiguities caused by
linear dependence amongst manifold vectors.
l can be ranging from 0 to 4 . The corresponding virtual
Next, if we consider the manifold curve's length and
sensor array can be obtained by Eq. (15). the rate-of-change of arc length, it is seen that as the
Using CRB as our performance indicator, the result is order of statistic increases, the lengths of curves are
shown in Fig. 2 illustrating the CRB as a function of increased from 4.44 to 8.89 ,14.05,19.87 , and 48.67
azimuth direction for different arrangement l . It is
observed that the array performance is maximized when respectively for the 4th , 6th ,8th and 16th order.
the l is equal to q / 2 (in case of q being an even And, since the curve length is longer, the
number). This result further confirms with the finding in corresponding rate-of-change of arc length se and
[8] that the array processing is optimized when the the arc length separation s between two directions are
integer l satisfies the condition in Eq. (7). increased as well.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
48
Supawat Supakwong
Fig. 2. The CRB as a function of azimuth angle for 8th order of statistic, plotted for different arrangement l
TABLE I
VIRTUAL SENSOR ARRAYS CORRESPONDING TO DIFFERENT ORDER OF STATISTIC 2q , THE TOTAL NUMBER
OF VIRTUAL SENSOR ELEMENTS NVS , AND THE ARRAY APERTURES
2 1 , 1 2 1
2 2
4 1, 0 ,1T 3 2
T
6 3 , 1 , 1 , 3 4 3
2 2 2 2
8 2 , 1, 0 ,1, 2T 5 4
TABLE II
IMPORTANT PARAMETERS IN CHARACTERIZING ARRAY MANIFOLD CURVES OF DIFFERENT ORDERS OF STATISTIC 2q
Order Arclength Rate-of-change Total curve length The principle curvature
2q se of arclength se lm i s
2 0.71 1 cos 0.71 sin 4.44 1.41
4 1.41 1 cos 1.41 sin 8.89 0.71
6 2.24 1 cos 2.24 sin 14.05 0.45
8 3.16 1 cos 3.16 sin 19.87 0.32
16 7.75 1 cos 7.75 sin 48.67 0.13
On the other hand, as the order of statistics increases, when the order of statistic increases. Given that the CRB
the fundamental curvature 1 s that essentially is inversely proportional to the square of the rate-of-
measures how fast the curve's angle is pulled away from change of manifold curve s , hence the bound is
the tangent line decreases. lowered as a result, as shown in Fig. 3(a).
This is as expected because with the higher-order In addition, the lower bounds at the array's broadside
statistics, the curve is longer in length, hence the rate-of- 90 are lower than at the array's endfire
change that the curve gets pulled away will decrease.
Analysis of the performance enhancement using ( 0 ,180 ) as already expected since it is easier to
higher-order statistics is now presented. As the order of process signal at broadside rather than at the array's
statistic increases, the lower bounds is lowered indicating endfire. Moreover, Figs. 3(b) and 3(c) show the array
the improvement on angular parameter estimation. performance enhancement in terms of the array's ultimate
Looking from the differential geometry perspective, this direction and resolution capabilities respectively.
improvement can be explained by considering the rate- The presented results are based on the analysis of the
of-change of arc length listed in Table II that is increased array detection and resolution in the presence of two
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
49
Supawat Supakwong
closely-located sources with an angle separation of Meanwhile, Fig. 4 shows the minimum SNR required
5. Both sources have equi-power, i.e. P1 P2 1. in order to detect or resolve two sources with an angular
Both indicators show similar patterns as in the case of separation of . The first source is impinging from the
CRB. This shows that, when using higher-order statistic, direction , which is chosen to be at 45 in this
the performance in terms of detection and resolution are example, and the second source is from the direction
also enhanced. In addition, note that the detection and . The number of snapshots is assumed to be
resolution thresholds with respect to the second-order L 10. It is seen that detecting the presence of sources is
statistics q 1 cannot be found. This is due to the fact easier than resolving sources.
that with the conventional second-order statistics an Furthermore, detecting/resolving sources that are
sensor array of two sensors cannot process more than a closely located (small ) requires a higher SNR as
single signal. compared to a scenario when sources are wider apart.
(a)
(b)
(c)
Figs. 3. The array performance as a function of azimuth angle, plotted for different orders of statistic (a) CRB (b) The detection's lower bound
in terms of SNR1 L (c) The resolution's lower bound in terms of SNR1 L
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
50
Supawat Supakwong
Fig. 4. The detection's and resolution's lower bounds in terms of SNR1 as a function of azimuth angle separation ,
plotted for different orders of statistic. L 10 , 45
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
51
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – The potential of electric power generation from marine renewable energy is
enormous. Ocean waves are being recognized as a resource to be exploited for the sustainable
generation of electrical power. The high load factors resulting from the fluid properties and the
predictable resource characteristics make ocean waves particularly attractive for power
generation and advantageous when compared to other renewable energies. Regarding this
emerging and promising area of research, this paper presents a complete review of wave energy
technologies describing, analyzing and fixing many of the concepts behind wave energy
conversion. The proposed review will specifically highlights the main wave energy conversion
projects around the world at different levels (demonstration stage, in production, and
commercialized projects). In particular, mooring will be discussed, as it is a key feature behind
massive deployment of wave energy converters. Finally, a discussion will highlight challenges that
wave energy converters need to overcome to become commercially competitive in the global
energy market. Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved.
52
H. Titah-Benbouzid, M. E. H. Benbouzid
1
Pw _ mcl g 2 H 2T (2)
32
Fig. 2. Global annual mean wave power estimation in kW/m spanning 10 years period [10]
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
53
H. Titah-Benbouzid, M. E. H. Benbouzid
OSCILLATING OSCILLATING
OVERTOPPING
WAVE SURGE WATER COLUMN
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
54
H. Titah-Benbouzid, M. E. H. Benbouzid
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
55
H. Titah-Benbouzid, M. E. H. Benbouzid
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
56
H. Titah-Benbouzid, M. E. H. Benbouzid
Mooring Anchor
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
57
H. Titah-Benbouzid, M. E. H. Benbouzid
It is therefore obvious that mooring design is a critical at first, small machines where developed first, and were
part of a WEC project. subsequently scaled-up to larger sizes and powers for
The devices are generally thought to be used in areas massive deployment.
of demanding environmental loads due to waves, current In fact, optimal wave energy absorption involves some
and wind. These survivability issues are addressed in kind of resonance. This implies that WECs geometry and
existing offshore standards, such as the DNV-OSE301 size are linked to wavelength.
[53]. So, if pilot plants are to be tested in the open ocean,
they must be large structures [6]. In this specific context,
challenges that WECs should to overcome to become
VI. Challenges for Commercial Viability commercially competitive leading to massive
It has been proven that wave energy extraction is very deployment could be summarized as:
attractive as it is spatially more concentrated than both – As for offshore converters, WECs should withstand
wind and solar energy; it is also more persistent and extreme wave condition leading to difficult and costly
predictable than wind energy. maintenance operations.
On the other hand, the development, from concept to – As above discussed, mooring design is a critical part.
commercial stage, has been found to be a very slow and In addition to the demanding environmental loads due
expensive process [11], [54]. Indeed, it is difficult to to waves, current and wind, the mooring system
follow what was done in the wind turbine industry where should also withstand constraints due to the WEC
alignment for capture optimization. Given the
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
58
H. Titah-Benbouzid, M. E. H. Benbouzid
continuous environmental loading, fatigue has been Proceedings of the 2014 IEEE PEAC, Shanghai (China), pp. 338-
342, November 2014.
identified as one of the key engineering challenges
[8] M.S. Lagoun, A. Benalia and M.E.H. Benbouzid, “Ocean wave
[55]. In addition, marine growth and corrosion need converters: State of the art and current status,” in Proceedings of
to be considered [56]. the 2010 IEEE ENERGYCON, Manama (Bahraïn), pp. 636-642,
– Higher costs of construction, deployment, and December 2010.
[9] Y. Hong, R. Waters, C. Boström, M. Eriksson, J. Engström and
maintenance need to be supported with substantial M. Leijon, “Review on electrical control strategies for wave
financial support from governments. energy converting systems,” Renewable and Sustainable Energy
Regarding mooring fatigue and cost issues, it has been Reviews, vol. 31, pp. 329-342, March 2014.
recently suggested to develop WECs without mooring [10] I. Lopez, J. Andreu, S. Ceballos, I. Martínez de Alegria and I.
Kortabarria, “Review of wave energy technologies and the
[57]-[58].
necessary power-equipment,” Renewable and Sustainable Energy
The developed converter has station-keeping ability Reviews, vol. 27, pp. 413-434, November 2013.
and evasive maneuver by diving. For station-keeping [11] B. Czech and P. Bauer, “Wave energy converter concepts: Design
within a uniform bound, a wave glider has been adopted challenges and classification,” IEEE Industrial Electronics
Magazine, vol. 6, n°2, pp. 4-16, June 2012.
as propulsion system [59] (Fig. 11).
[12] B. Drew, A.R. Plummer and M.N. Sahinkaya, “A review of wave
In addition, such kind of system has the ability to energy converter technology,” Proc. IMechE, Part A: Journal
submerge to a certain depth for its safety in emergency Power and Energy, vol. 223, n°8, pp. 887-902, December 2009.
condition such as typhoon [60]. [13] J. Falness, “A review of wave-energy extraction,” Marine
Structures, vol. 20, n°4, pp. 185-201, October 2007.
[14] S. Olaya, J.M. Bourgeot and M.E.H. Benbouzid, “Hydrodynamic
coefficient computation for a partially submerged wave energy
converter,” IEEE Journal of Oceanic Engineering, vol. PP, n°99,
pp. 1-14, 2014.
[15] S. Olaya, J.M. Bourgeot and M.E.H. Benbouzid, “Optimal control
for a self-reacting point absorber: A one-body equivalent model
approach,” in Proceedings of the 2014 IEEE PEAC, Shanghai
(China), pp. 332-337, November 2014.
[16] S. Olaya, J.M. Bourgeot and M.E.H. Benbouzid, “Modelling and
preliminary studies for a self-reacting point absorber WEC,” in
Proceedings of the 2014 IEEE ICGE, Sfax (Tunisia), pp. 14-19,
March 2014.
[17] K. Rhinefrank et al., “Comparison of direct-drive power takeoff
Fig. 11. A mooring-less wave energy converter concept [57] systems for ocean wave energy applications,” IEEE Journal of
Oceanic Engineering, vol. 37, n°1, pp. 35-44, January 2012.
[18] T.K.A. Brekken, H.M. Hapke, C. Stillinger and J. Prudell,
“Machines and drives comparison for low-power renewable
VII. Conclusion energy and oscillating applications,” IEEE Trans. Energy
This paper has proposed an up-to-date review of the Conversion, vol. 25, n°4, pp. 1162-1170, December 2010.
[19] L. Cappelli, F. Marignetti, G. Mattiazzo, E. Giorcelli, G. Bracco,
most recent trends regarding main wave energy converter S. Carbone, C. Attaianese, “Linear tubular permanent-magnet
technologies with respect to overviews already published generators for the inertial sea wave energy converter,” IEEE
in the past years. In addition, mooring has been discussed Trans. Industry Applications, vol. 50, n°3, pp. 1817-1828,
and has been shown to be a key feature behind massive May/June 2014.
[20] C. Boström, B. Ekergard, R. Waters, M. Eriksson and M. Leijon,
deployment of wave energy converters. Finally it has “Linear generator connected to a resonance-rectifier circuit,”
been highlighted some challenges that needs to be IEEE Journal of Oceanic Engineering, vol. 38, n°2, pp. 255-262,
overcome to enlarge the vision of large-scale commercial April 2013.
arrays of wave energy converters. [21] T.K.A. Brekken, K. Rhinefrank, A. von Jouanne, A. Schacher, J.
Prudell and E. Hammagren, “Scaled development of a novel wave
energy converter including numerical analysis and high-resolution
tank testing,” Proc. IEEE, vol. 101, n°4, pp. 866-875, April 2013.
References [22] R. Vermaak and M.J. Kamper, “Design aspects of a novel
topology air-cored permanent magnet linear generator for direct
[1] J. Crus Ed., Ocean Wave Energy: Current Status and Future drive wave energy converters,” IEEE Trans. Industrial
Perspectives. Springer, Berlin, 2008. Electronics, vol. 59, n°5, pp. 2104-2115, May 2012.
[2] J.G. Vining and A. Muetze, “Economic factors and incentives for [23] N. Hodgins, O. Keysan, A.S. McDonald and M.A. Mueller,
ocean wave energy conversion,” IEEE Trans. Industry “Design and testing of a linear generator for wave-energy
Applications, vol. 45, n°2, pp. 547-554, March-April 2009. applications,” IEEE Trans. Industrial Electronics, vol. 59, n°5,
[3] Z. Zhou, F. Scuiller, J.F. Charpentier, M.E.H. Benbouzid and T. pp. 2094-2103, May 2012.
Tang, “An up-to-date review of large marine tidal current turbine [24] J. Zhang, H. Yu, Q. Chen, M. Hu, L. Huang and Q. Liu, “Design
technologies,” in Proceedings of the 2014 IEEE PEAC, Shanghai and experimental analysis of AC linear generator with Halbach
(China), pp. 448-484, November 2014. PM arrays for direct-drive wave energy conversion,” IEEE Trans.
[4] S. Benelghali, M.E.H. Benbouzid and J.F. Charpentier, “Marine Applied Superconductivity, vol. 24, n°3, #0502704, June 2014.
tidal current electric power generation technology: State of the art [25] R. Ekström, B. Ekergård and M. Leijon, “Electrical damping of
and current status,” in Proceedings of the 2007 IEEE IEMDC, linear generators for wave energy converters—A review,”
Antalya (Turkey), vol. 2, pp. 1407-1412, May 2007. Renewable and Sustainable Energy Reviews, vol. 42, pp. 116-128,
[5] D. Ross, Power from the Waves. Oxford University Press: 1995. February 2015.
[6] F.A.O. Falcao, “Wave energy utilization: A review of the [26] H. Polinder, B.C. Mecrow, A.G. Jack, P.G. Dickinson and M.A.
technologies,” Renewable and Sustainable Energy Reviews, vol. Mueller, “Conventional and TFPM linear generators for direct-
14, n°3, pp. 899–918, April 2010. drive wave energy conversion,” IEEE Trans. Energy Conversion,
[7] H. Titah-Benbouzid and M.E.H. Benbouzid, “Ocean wave energy vol. 20, no. 2, pp. 260-267, June 2005.
extraction: Up-to-date technologies review and evaluation,” in
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
59
H. Titah-Benbouzid, M. E. H. Benbouzid
[27] M.A. Mueller, “Electrical generators for direct drive wave energy floating WEC devices,” Ocean Engineering, vol. 34,n°14-15, pp.
converters,” IEE Proc. Generation, Transmission & Distribution, 1918-1934, 2007.
vol. 149, n°4, pp. 446-456, July 2002. [47] Z. Gao and T. Moan, “Mooring system analysis of multiple wave
[28] N. Muller, S. Kouro, J. Glaria and M. Malinowski, “Medium- energy converters in a farm configuration,” in Proceedings of the
voltage power converter interface for WaveDragon wave energy 2008 EWTEC, Uppsala (Sweden), pp. 509-518, September 2009.
conversion system,” in Proceedings of the IEEE ECCE, Denver [48] A. Pecher, A. Foglia and J.P. Kofoed, “Comparison and sensitivity
(USA), pp. 352-358, September 2013. investigations of a CALM and SALM type mooring system for
[29] M.S. Lagoun, A. Benalia, and M.E.H. Benbouzid, “A predictive wave energy converters,” Journal of Marine Science and
power control of doubly fed induction generator for wave energy Engineering, vol. 2, pp. 93-122, 2014.
Converter in irregular waves,” in Proceedings of the 2014 IEEE [49] H. Ming, and G. A. Aggidis, “Developments, expectations of
ICGE, Sfax (Tunisia), pp. 26-31, March 2014. wave energy converters and mooring anchors in the UK,” Journal
[30] M.S. Lagoun, S. Nadjem, A. Benalia and M.E.H. Benbouzid, of Ocean University of China, vol. 7, n°1, pp. 10-16, February
“Predictive power control of doubly-fed induction generator for 2008.
wave energy converters,” in Proceedings of the 2012 EFEA [50] F. Cerveira, N. Fonseca and R. Pascoal, “Mooring system
(International Symposium on Environment Friendly Energies and influence on the efficiency of wave energy converters,”
Applications), Newcastle upon Tyne (UK), pp. 312-317, June 2012. International Journal of Marine Energy, vol. 3-4, pp. 65-81,
[31] M. Alberdi, M. Amundarain, A.J. Garrido and I. Garrido, “Neural December 2013.
control for voltage dips ride-through of oscillating water column- [51] B. Zanuttigh, E. Angelelli and J.P. Kofoed, “Effects of mooring
based wave energy converter equipped with doubly-fed induction systems on the performance of a wave activated body energy
generator,” Renewable Energy, vol. 48, pp. 16-26, 2012. converter,” Renewable Energy, vol. 57, pp. 422-431, September
[32] M. Alberdi, M. Amundarain, A.J. Garrido, I. Garrido and F.J. 2013.
Maseda, “Fault-ride-through capability of oscillating-water- [52] W. Valentine and K.D. von Ellenrieder, “Model scaling of ocean
column-based wave-power-generation plants equipped with hydrokinetic renewable energy systems,” IEEE Journal of
doubly fed induction generator and airflow control,” IEEE Trans. Oceanic Engineering, vol. 1, n°1, pp. 27-36, January 2015.
Industrial Electronics, vol. 58, n°5, pp. 1501-1517, May 2011. [53] Det Norske Veritas offshore standard DNV-OS-E301, Position
[33] S. Benelghali, M.E.H. Benbouzid and J.F. Charpentier, “Generator Mooring, October 2010.
systems for marine current turbine applications: A comparative [54] G. Beaudoin, D. Robertson, R. Doherty, D. Corren, B. Staby and
study,” IEEE Journal of Oceanic Engineering, vol. 37, n°3, pp. L. Meyer, “Technological challenges to commercial-scale
554-563, July 2012. application of marine renewables,” Oceanography, vol. 23, n°2,
[34] E.V Sanchez, R.H. Hansen and M.M. Kramer, “Control pp. 32-41, June 2010.
performance assessment and design of optimal control to harvest [55] P.R. Thies, L. Johanning, V. Harnois, H.C.M. Smith and D.N.
ocean energy,” IEEE Oceanic Engineering, vol. 40, n°1, pp. 15- Parish, “Mooring line fatigue damage evaluation for floating
26, January 2015. marine energy converters: Field measurements and prediction,”
[35] J. Aubry, H. B. Ahmed, B. Multon, “Sizing optimization Renewable Energy, vol. 63, pp. 133-144, March 2014.
methodology of a surface permanent magnet machine-converter [56] B. Polagye and J. Thomson, “Screening for biofouling &
system over a torque-speed operating profile: application to a corrosion of tidal energy device materials,” Northwest National
wave energy converter,” IEEE Trans. Industrial Electronics, vol. Marine Renewable Energy Center, Oregon State University,
59, n°5, pp. 2116-2125, May 2012. Report 1, April 2010.
[36] A. Babarit, J. Hals, M.J. Muliawan, A. Kurniawan, T. Moan, J. [57] H. Joe, M. Kim, S.M. Wi, H.S. Kwon and S.C. Yu, “Development
Krokstad, “Numerical benchmarking study of a selection of wave of mooring-less robotic buoy system using wave powered
energy converters,” Renewable Energy, vol. 41, pp. 44-63, May renewable energy,” in Proceedings of the 2014 MTS/IEEE
2012. OCEANS, Newfoundland (Canada), pp. 1-6, September 2014.
[37] L. Margheritini, D. Vicinanza and P. Frigaard, “SSG wave energy [58] M. Kim, S.M. Wi, H. Joe, H.S. Kwon and S.C. Yu, “Multi-body
converter: Design, reliability and hydraulic performance of an point absorber system without a mooring,” in Proceedings of the
innovative overtopping device,” Renewable Energy, vol. 34, n°5, 2014 MTS/IEEE OCEANS, Taipei (Taiwan), pp. 1-4, April 2014.
pp. 1371-1380, May 2009. [59] R. Hine, S. Willcox, G. Hine and T. Richardson, “The Wave
[38] Vryhof Anchors: http://www.vryhof.com/ammanual.htm Glider: A Powered autonomous marine vehicle,” in Proceedings
[39] I.M.L. Ridge, S.J. Banfield and J. Mackay, “Nylon fibre rope of the 2009 MTS/IEEE OCEANS, Biloxi (USA), pp. 1-6, October
moorings for wave energy converters,” in Proceedings of the 2014.
2010 MTS/IEEE OCEANS, Seattle (USA), pp. 1-10, September [60] J.S. Lee and S.C. Yu, “Preliminary study of long endurance moor-
2010. less non-drifting buoy,” in Proceedings of the 2013 IEEE IUTS,
[40] J.D. Pasternak, M. Hersley and S. Leite, “Increased offshore Tokyo (Japan), pp. 1-6, March 2013.
MODU mooring performance through MODUline® polyester
rope,” in Proceedings of the 2010 MTS/IEEE OCEANS, Seattle
(USA), pp. 1-8, September 2010. Authors’ information
[41] R.E. Harris, L. Johanning and J. D. Wolfram, “Mooring systems
for wave energy converters: a review of design issues and University of Brest, EA 4325 LBMS, Rue de Kergoat, CS 93837, 29238
choices,” in Proceedings of the 2004 MAREC, Blyth (UK), pp. 1- Brest Cedex 03, France.
10, July 2004.
[42] Bruce Connector: http:// www.bruceanchor.co.uk/ Hosna Titah-Benbouzid was born in Annaba,
[43] P.R. Thies, L. Johanning and P. McEvoy, “A novel mooring tether Algeria, in 1973. She received the Engineer
for peak load mitigation: Initial performance and service degree in agro-alimentary engineering from the
simulation testing,” International Journal of Marine Energy, vol. University of Constantine, Constantine, in 1998,
7, pp. 43-56, September 2014. the M.Sc. degree in environment engineering
[44] M. Tavakoli, T.K.A. Brekken, B. Bosma and A.A. Schacher, from the University of Picardie, Amiens, France,
“programmable mooring controller for tank testing of scaled wave in 2006, and the PhD degree in Chemical and
energy converters,” in Proceedings of the 2013 IEEE PES Environmental engineering from the University
General Meeting, Vancouver (Canada), pp. 1-5, July 2013. of Brest, Brest, France, in 2010. Dr. Titah-Benbouzid is an affiliate
[45] L. Johanning, G.H. Smith and J. Wolfram, “Mooring design member of the LBMS_Lab (EA 4325) since 2012. Her current research
approach for wave energy converters,” Proc. Inst. Mech. Eng., interests are marine renewable energy systems interactions with marine
Part M: Journal of Engineering for the Maritime Environment, environment and biofouling.
vol. 220, n°4, pp. 159-174, December 2006. E-mail: hosna.titahbenbouzid@gmail.com
[46] L. Johanning, G.H. Smith and J. Wolfram, “Measurements of
static and dynamic mooring line damping and their importance for
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
60
H. Titah-Benbouzid, M. E. H. Benbouzid
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
61
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – Low-voltage distribution networks represent a component of the power system with
the most prominent problems in terms of voltage quality and losses of electrical power which is
delivering to the end customers. Overcoming of such problems is a real challenge for distribution
companies, since they are obliged to deliver electricity of standard quality to the end customers,
but at the same time taking care of their own interests. In this paper, a method based on
hierarchical algorithms applicable in the planning of low-voltage distribution networks is
presented. The planning of such networks includes determining of an optimum location of
substation, radial network configuration, as well as appropriate types of power lines, all in order
to achieve the prescribed voltage levels and minimize the power losses in electricity transmission.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved.
62
Amir Softic, Marinko Stojkov, Admir Jahic
to solve this technical problem can be the use of modern This distance corresponds to the length of the line
hierarchical algorithms. which connects elements 1 and 2. In the same way, the
distances between all pairs of elements can be calculated.
All these distances are usually presented in the form of
II. Background Theory of Applied Models a matrix of distances. In this matrix, nondiagonal
In the literature, applied methods can be found under elements represent the distances between the pairs, and
different names, but by their nature, they can be diagonal elements are 0 (the distance of element from
classified into hierarchical methods. In the basis of these itself). Since the distance between objects 1 and 2 is
methods lies iterative process of connecting objects into identical to the distance between the objects 2 and 1, the
the groups [5]. The proposed model of planning of low- distance matrix is symmetric. This matrix is presented in
voltage distribution networks uses two hierarchical Table I.
methods, namely: Fuzzy clustering (FC) and Analytic
TABLE I
Hierarchy Process (AHP). EUCLIDEAN MATRIX OF DISTANCES
Objects 1 2 3 4 5 6 7
1 0 , , , , , ,
II.1. Fuzzy Clustering
2 , 0 , , , , ,
Fuzzy clustering is a method by which the appropriate 3 , , 0 , , , ,
1
( )=
( . )
(1)
∑
,
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
63
Amir Softic, Marinko Stojkov, Admir Jahic
II.2. Analytic Hierarchy Process The first step in AHP procedure is making
comparisons between pairs of each criterion. Any
Analytic Hierarchy Process (AHP) was developed by
decision can be presented with number at a certain scale.
Thomas L. Saaty, university professor (University of
One commonly used scale is presented in Table II [9].
Pittsburgh), in 70's and 80's of the 20th century and often,
in the literature, can be found under the name Saaty's TABLE II
method. This method allows the user to assess the SAATY’S RATING SCALE
relative weight of multiple criteria or multiple options in Intensity
of Definition Explanation
relation to the given criterion, in an intuitive way. In case
importance
those quantitative assessments are not available, the 9 Extreme importance The evidence favouring
assessor still can recognize whether some criterion is one over the other activity
more important than another. Satty has established a is of the highest possible
consistent way of converting the comparisons (X is more order of affirmation
important than Y) in a set of numbers that represents the 7 Very strong or demonstrated An activity is favoured
relative priority of each criterion. The idea of the AHP importance very strongly over another;
its dominance
method is, at first, to establish the goal, and then to demonstrated in practice
establish the criteria which can satisfy and also criteria
5 Strong importance Experience and judgement
which has the influence on the decision of the decision strongly favour one activity
maker. Criteria can be classified into one or more levels over another
(first, second, ...) which provides a hierarchy of criteria, 3 Moderate importance Experience and judgement
whereby the criteria of the first level have the greatest slightly favour one activity
influence on the decision (Fig. 2). over another
The basic of applying AHP consists of next steps: 1 Equal importance Two activities contribute
Structuring of decision problem and selection of equally to the objective
criteria, 2,4,6,8 Intermediate values When compromise is
Establishing of criteria priorities by comparing of needed
pairs (weighing),
Comparison of pairs of options for each criterion The results of comparison(for each pair) are given as
(scoring), integer values from 9 (extremely different) to 1 (equal
Obtaining an overview of the relative scoring for each value), where a larger number means that the selected
option. factor is observed as more important, in greater degree
At first, the AHP method can be implemented in four than the second factor with which it is compared.
basic steps: Weighted ratios are determined by equation:
Decomposition of the problem into hierarchy,
Comparison by pairs and establishing of priorities = (5)
among the elements in the hierarchy,
Synthesizing of decisions for getting a set of overall
decisions or weights to achieve the goal, where , are relative importances of the criteria.
Evaluation and verification of consistency of The weighted ratios are recorded in the form of a
decisions. weighted ratios matrix, in the form:
The basic procedure consists from: ⋯
Developing a score for each decision, alternative for
= ⋮ ⋱ ⋮ (6)
each criterion,
⋯
Developing the weights for criteria,
Calculation of weighted average assessments for each where q is the number of criterion.
alternative decision. Selecting the one with highest
score.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
64
Amir Softic, Marinko Stojkov, Admir Jahic
= (10)
TABLE III
RANDOM CONSISTENCY INDEX
n 1 2 3 4 5 6 7 8 9 10
RI 0 0 0.58 0.90 1.12 1.24 1.32 1.41 1.45 1.49
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
65
Amir Softic, Marinko Stojkov, Admir Jahic
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
66
Amir Softic, Marinko Stojkov, Admir Jahic
This method formulates alternatives and ranks certain These data we were
re necessary for calculation of the
criteria, and based on them it shows which of offered voltage drops, power losses and power lines loads. The
alternatives is the best. Since the ranking ooff criteria is data for assessment of overall costs, needed fo forr
issue of subjective assessment of someone who performs construction of the network, we were
re also obtained.
the planning
planning,, that can lead to the error in the ranking of Calculation of other rrequired
equired data was
was performed in
criteria, so this step has to be repeated until it satisfied Matlab (Matpower) [11]. For the calculation of voltagevoltage
the condition of consistency. drops, power losses and
and power lines loads are used the
When this condition is satisfied
satisfied as tthe
he final solution,
solution data of installed capacity of substation, customer’s loads
an optimal location and installed capacity of substation, and parameters of power lines
line sections ((r, x,
x, length). The
The
the optimal number of outages and topological obtained data about voltage drops, for each alternative,
configuration of low-low-voltage
voltage distribution network are are presented in Tabl
Tablee IV.
obtained.
obtained
TABLE IV
CALCULATED VALUES OF VOLTAGE DROPS
Number of Un ΔU U ΔU ΔU
IV. Results clusters (V) (p.u) (V) (V) (%)
The functionality of the proposed methodology is 5 400 0.930 372.0 -28.0
28.0 -7.0
7.0
6 400 0.934 373.6 -26.4
26.4 -6.6
6.6
tested using the example of a real low low--voltage
voltage 7 400 0.935 374.0 -26.0
26.0 -6.5
6.5
distribution network. It supplies 270 70 electricity 8 400 0.940 376.0 -24.0
24.0 -6.0
6.0
customers
customers, in the category of households
households, through
hrough this
network. The network is powered by the transformer The obtained data of power losses, for each
station 10(20)
10(20)/0..4
4 kV,
kV rated power of 400 kVA. alternative, are presented in Table V.
Customers loads w were
re calculated using method of
peak loads based on annual realized consumption of each TABLE V
customer and daily, one-hour
one hour measurements of low-
low CALCULATED VALUES
LUES OF POWER LOSSES
voltage outages loads (Fig. 8). Number of clusters Losses (kW) Losses (%)
5 7.934 3.477
6 7.114 3.118
7 7.098 3.111
8 6.581 2.844
TABLE VI
AVERAGE LOADS OF LINES
INE
Number of In I average I average
clusters (A) (A) (%)
5 198 71.62 36.17
6 198 59.48 30.04
7 198 50.98 25.75
8 198 44.51 22.48
Fig. 8. Daily load curve of TS 10(20)/0
10(20)/0.4
4 kV, 400 kVA
Another data required for the application of the AHP
At ffirst
irst,, a network configura tion according tto
configuration o the
method are estimated costs. Table VII presents an
proposed
oposed methodology was made and the calculations
overview of overall costs for each alternative
alternative.
were car
were ried out, and then the results we
carried were
re compared
with the data obtained for the existing network. Thus, TABLE VII
according to the methodology, at first, the coordinates of OVERALL ESTIMATED COSTS
all measurement points were obtained from Number of clusters Costs ((€)
georeferenced maps.
maps They represent
represented the input data for 5 207395
6 238631
the application of fuzzy algorithm. Using this algorithm, 7 263265
in first step, it was defined a geograp
geographic
hic location of 8 268912
substation
substation. After that, using the same algorithm, a
number anandd spatial distribution of customers groups were After calculation all of necenecessary
ssary data, the final
determined
determined, which also represent
representeded a number of low-
low solution wa
wass obtained by applying the AHP method. Four
Four
voltage outages. Since,
ince, for obtaining a final solution wawas alternatives (networks
(networks with 5, 6, 7 and 8 lowlow-voltage
voltage
necessary to have more aalternatives,
lternatives, four alternatives outages)are
outages) are obtained by Fuzzy algorithm.
algorithm. After that, the
were determined. Alternatives we re formed oon
were n the basis criteria we
were
re determined, namely:
of the number of low low-voltage
voltage outages (5, 6, 7 and 8 Criterion of voltage drops,
outages). After that, in AutoCAD, a drawing of low- low Criterion of power losses,
voltage networks was donedone,, for all formed alternatives.
alternatives Criterion of average lines loads,
This was provided the network data. Criterion of overall costs.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10,, N. 1
67
Amir Softic, Marinko Stojkov, Admir Jahic
Ranking of criteria was carried out at first. In first step TABLE XIII
CALCULATION OF PRIORITIES IN RELATION TO ALTERNATIVES
a pair wise comparison matrix was established, as it is
Costs Alt. 1 Alt. 2 Alt. 3 Alt. 4 Priority
presented in Table VIII. Alt. 1 1 1.1506 1.2694 1.2966 0.2917
Alt. 2 0.8691 1 1.1032 1.1269 0.2535
TABLE VIII Alt. 3 0.7878 0.9064 1 1.0215 0.2298
PAIR WISE COMPARISON MATRIX Alt. 4 0.7712 0.8874 0.9790 1 0.2250
Voltage Lines
Losses Costs
drops loads TABLE XIV
(C2) (C4)
(C1) (C3) WEIGHTED AVERAGE RATINGS FOR EACH ALTERNATIVE
Voltage drops 1 2 4 5 Alternative Weighted average ratings
Losses 0.50 1 4 5 Alternative 1 0.24154
Loads of lines 0.25 0.25 1 2 Alternative 2 0.24776
Costs 0.20 0.20 0.50 1 Alternative 3 0.24682
Alternative 4 0.26390
After determining the rows weights and priorities
values of comparison matrix there were determined the From these results, it is obvious that alternative four is
vectors of eigenvalues of the comparison matrix as it is the best solution, as it is provided the highest value of the
presented in Table IX. vector of priorities. It means that the optimal solution for
such low-voltage distribution network is construction of
TABLE IX the network with eight outages. From the table, too, it
PRIORITY VECTORS can be concluded that a larger number of low-voltage
Priority outages does not necessarily mean a better solution
C1 C2 C3 C4 Ʃ
vectors because the solution with six outages, in this case, is
C1 0.5128 0.5797 0.4211 0.3846 1.8982 0.4746
C2 0.2564 0.2899 0.4211 0.3846 1.3519 0.3380
better than solution with seven outages. This solution
C3 0.1282 0.0725 0.1053 0.1539 0.4598 0.1149 was compared with the existing distribution network in
C4 0.1026 0.0580 0.0526 0.0769 0.2901 0.0725 terms of voltage drops and power losses. The values of
voltage drops and power losses for the existing network,
After calculating = 4.08876, it was necessary to with five low-voltage outages, are calculated in
examine consistency. After calculating, the values Matpower and the results are presented in Table XV.
= 0.02959 and = 0.03287were obtained.
Since ≤ 0.1, it means that ranking of criteria was TABLE XV
VOLTAGE DROPS AND POWER LOSSES IN EXISTING NETWORK
performed well. After that, the calculation of priorities of
specific criteria in relation to alternatives was done. The Un ΔU U ΔU ΔU Losses Losses
(V) (p. u) (V) (V) (%) (kW) (%)
results are presented in Tables X – XIII.
After obtaining of these results, a multiplication of 400 0.909 363.6 -36.4 -9.1 7.226 3.167
these priorities with values of comparison matrix was
performed. The obtained results are presented in Table Comparative results with the solution obtained by
XIV. proposed methodology are presented in Table XVI.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
68
Amir Softic, Marinko Stojkov, Admir Jahic
It implies the planning process from selecting the type demands for the quality of supplied electricity have the
and location of the substation to the determination of the primary role. In addition, by defining the configuration of
optimal configuration of low-voltage network. distribution network, distribution companies can pre-plan
On a real example it is easy to see that methodology the necessary funds for the reconstruction or construction
gives better results in comparison with the existing of distribution networks. Therefore, this methodology
distribution network, especially in the case of voltage can be successfully applied to the reconstruction process
drops at end customers. This is particularly important and in the process of constructing a completely new low-
because in the process of market deregulation customers’ voltage distribution networks, while at the same time can
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
69
Amir Softic, Marinko Stojkov, Admir Jahic
also be used to detect problems in the network, as well as Marinko Stojkov (1970, Croatia) graduated
BSc (1994), MSc (1998) and PhD (2002) on
to offer suggestions for their solution. By correct ranking
Faculty of Electrical Engine
Engineering,
ering, Energy
of criteria and the precise definition of the network Department, University of Zagreb. He was
topology,, methodology gives better results.
topology employed in HEP HEP-DSO
DSO (1995-2009)
(1995 2009) on
Therefore, for an experienced engineers who positions: manager at the Maintenance
Department, Head of Department for
participate in the planning process this methodology can Development and Head of Planning and
serve resented
rve as a very useful tool. Presen
Presen ted methodology,
methodology in the Investment of Elektra SlavonskiBrod. He started his academy career on
same way,
way can be used for planning the medium
medium--voltage
voltage Faculty of Electrical Engineering, University of Osijek (1998). He was
distribution networks. employed as Assistant Professor on Mechanical Engineering Faculty in
SlavonskiBrod, University of Osijek (2009) and as Associate professor
(2010) in the scientific field of technical sciences - electrical
engineering. He attended LV and MV cables and cable equipment
References course (1996) in Budapest, Hungary and "Electric Distribution
[1] Eloy Diaz-Dorado,
Diaz Dorado, Jose Cidras Pidre, Edelmiro Miguez Garcia, Garcia Management" course (2003) in Dublin, Republic of Ireland. He
Planning of Large Rural LowLow-Voltage
Voltage Networks Using Evolution published 3 CC scientific journal papers, 5 SCI SCI-Expanded
Expanded journal
Strategies, IEEE T
Strategies, Transactions
ransactions on Power Systems, vol. 18, No. 44, scientific papers, 8 papers in other scie
scientific
ntific journals, 25 international
November 2002003. scientific conferences papers, 1 invited lecture, 15 expert papers. Also
[2] Alejandro Navarro, Hugh Rudnick, Large Large-Scale
Scale Distribution he reviewed 20 scientific papers in journals and 4 international
Planning Part I: Simultaneous Network and Transformer
Planning-Part conferences papers.His research interests are: power distribution
Optimization, IEEE Transactions on Power Systems, vol. 24, No. network, power loslosses,
ses, harmonic distortion, renewable energy sources,
2,, May 2009. load forecast, electric and magnetic fields, energy efficiency and energy
[3] Zvonimir Klaic, Srete Nikolovski, Zorislav Kraus Kraus,, Voltage savings. He is a member of IEEE.
variation performance indices in distribution network
network,, Technical
Gazette, Vol. 18
18,, No. 4, 547 551, 2011.
4 pp. 547-551, Admir Jahic was born in Tuzla in Bosnia and
[4] Ian Dent, Tony Craig, Uwe Aickelin, Tom Rodden Rodden,, An Approach Herzegovina, on December 4, 1975. He received
for Assessing Clustering
Clustering of Households by Electricity Usage Usage, his gr
graduation
aduation (dipl. Eng.) and M.Sc. degree
2014.
Cornell University, September 2014 from the Faculty ofElectrical Engineering at the
[5] Sadeghi Sarcheshmah, M., Seifi, A.R., A new fuzzy power flow University of Tuzla, both in electrical
analysis based on uncertain inputs, (2009) International Review of engineering. In 2011 elected for a teaching
Electrical Engineering (IREE), 4 (1), pp. 122122-128
128. assistant in the Department of Power Systems at
[6] Paracha, Z.J., Kalam, A., F Fuzzy
uzzy clustering techniques for the Faculty of Electrical Engineering Tu Tuzla.
zla.
analysis of PQ data in electrical power distribution system, (2010) Currently holds a position of an “Engineer for the energetics and
International Review of Electrical Engineering (IREE), 5 (5), pp. investments” in “Elektrodistribucija“ Tuzla, a distributive part of a
2152 2158.
2152-2158. Public Enterprise Elektroprivreda B&H – Sarajevo, the public
[7] Marinko Stojkov, MirzaAtic, Amir Softic, Application of Fuzzy enterprise for generation, distributi
distribution
on and sale of electrical power. His
logic
ogic for reactive power compensation by synchronous motors, topics of interest are
are: analysis, optimization, planning, control, state
Technical Gazette, Vol.19, No. 4, pp. 753
753-758,
758, 2012. estimation and distribution networks automation.He is a member of
[8] Eduardo Gouveia, Manuel A. Matos Matos,, Constrained Fuzzy Power IEEE, CIGRE and CIRED.
Flow
Flow,, Power Tech, IEEE Lausanne,
Lausanne, 2007
2007.
[9] Thomas L. Satty, Decisions making with the analytic hierarchy
process, Int. J. Services Sciences, vol. 1, No. 11,, 2008
2008.
[10] Thomas L. Satty, How to make a decision: The Analytic
[10
Hierarchy Process, Journal of Operational Research
Research, vol. 24,Issue
6, December 1994.
[11 Ray D. Zimmerman, Carlos E. Murillo-Sanchez,
[11] Murillo Sanchez, MATPOWER
4.1 User's Manual, 2011.
Authors’ information
Amir Softic
Softi was born in Tuzla (Bosnia and
Herzegovina) on 30th of January, 1964. He
Herzegovina)
received a B.S. degree in Electrical Engineering
and M.S. de degree
gree from University of Tuzla
(Bosnia and Herzegovina
Herzegovina),), in 1996 and 2009,
respectively. From 2005 he is working in PE
ElektroprivredaBiH Sarajevo, at works which
are related to distribution networks analysis and
management.His topics of interest are are: distribution networks, power
network optimization,
optimization, energy analysis and distribution network
planning, supervising and managing
managing,, electromagnetic compatibility,
renewable energy sources.
sources
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10,, N. 1
70
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – In this paper the one possibility of single phase fault location in unearthed MV
distribution network is presented. This method is based on transients which appear in the case of
charge redistribution in the instance of fault occurrence. The issues of fault location in
distribution networks was emphasized through influence of inhomogeneity, distributed loads,
considerable fault resistance and numerous branches on distribution feeders. On the extracted
transients of currents and voltages of faulty phase, method based on solving differential equations
is applied.
The low frequency charge transients, with minimum sample frequency of several kHz is used. This
method was examined by computer simulations on the modeled network as well as on the real
network and fault. In the case of real fault, it was used data registered by power quality analyzer.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved.
Keywords: Distribution Medium Voltage (MV) Network, Fault Location, Charge Transients
Nomenclature I. Introduction
vA, vB, vC Transient voltages of phases A, B and C [V] For this type of fault location the currents and voltages
iA, iB, iC Transient currents of phases A, B and C [A] at the sending end of distribution feeder in supplied
v0 Zero sequence voltage [V] transformer station (TS) registered during a fault are
i0 Zero sequence current [A] used. This way of determining fault place is referred as
t Time [s] on-line location, automated or fault location in real time.
x Estimated distance to the fault place [km] In transmission networks of voltage levels 110, 220
Δt Sampling time [s] and 400 kV is in almost regular usage. In medium
Rf Fault resistance [Ω] voltage (MV) distribution networks, commercial
Cz, Cm Ground and phase to phase feeder modules of fault locators occur only in recent years.
capacitances [µF] These are mainly fault locators that estimate fault
Ltr, Lv Transformer and feeder inductances [H] location on the basis of fundamental frequency currents
ω Angular frequency [rad/s] and voltages and reactance estimation to the fault place.
vk Phase voltage at fault time inception [V] If a network is branched it can be happened that
Iz Uncompensated RMS value of stationary estimated location corresponds to multiple points in a
fault current [A] network (problem of multiple fault location) or, if a
Ruz Resistance of neutral transformer grounding neutral point of network is isolated, there is no possibility
[Ω] of determining a place of single-phase short circuit.
l Total length of a feeder [km] Regarding that a single-phase short circuit is the most
Rx Faulty path resistance [Ω] common fault type (approximately 70% of faults in
Lx Faulty path inductance [H] distribution networks are single-phase faults), usability
L0’, L1’, L2’ Feeder inductances of zero, positive and of these devices is small in isolated network and in these
negative sequences per unit length [H/km] networks they can be used only for location phase to
Ik3 Short circuit current [A] phase faults. In grounded network, it is possible to
f Frequency [Hz] estimate impedance to a fault place also in the case of
ϑ Phase angle [°] single-phase short circuit. Difficulties with determining
x1 , x0 Positive and zero sequence reactances of a reliable location of single-phase short circuit in the case
feeder [Ω] of MV distribution networks with isolated neutral point,
Ceq Equivalent capacity [µF] are caused by the fact that fundamental frequency fault
Leq Equivalent inductance [H] current often small compared with phase to phase fault
I Initial amplitude of a charge transient [A] currents. This happens even if fault resistances are small,
i.e. in the case of direct single-phase short circuit (or
grounding).
71
Š. Gruhonjić Ferhatbegović
It should be borne in mind that a large portion of MV After identification of the nearest reactance value, it
network works with isolated neutral point. This applies follows determination of a distance to the fault place in
particularly to overhead network, which are often the [km] from nearest nodal point or nearest 10(20)/0,4 kV.
longest and with the most number of branches [1], [2]. In this way a location where team for repairing and
The choice of direct transformation 110/20 kV, may remedy a problem need to go out is determined. The
further highlight the need and advantages of automated problem of fault locating in distribution networks was
fault location. The rated voltage of 20 kV, according to considered in the literature [4]-[8].
the characteristics of topology, configuration, neutral
point treatment, the existence of branches and distributed
loads remains the same as in distribution feeders at 10 II. Charge Transient Frequency
kV. The difference is that 20 kV distribution feeders, Calculation
because of better transmission capacity, can supply
Transient fault location methods use currents and
customers in a wider area within allowed losses and
voltages in transient conditions that occur in a process of
voltage drops. This will cause longer feeders and
discharge and charge network capacity at the moment of
distance estimation during a fault, may be of great
single phase fault inception. In general, distribution
importance and can significantly reduce cost of field,
system is made up of resistance, inductance and capacity
value of unserved energy and periods without voltage.
which exist as concentrated parameters but are also
In relation to a fault location in transmission network,
distributed along the line, transformers, etc.. During each
in the case of fault location in MV network there are
state variation, such as operation of interrupting of short
certain characteristics that make difficult estimation of
circuit, fault condition, this system should generate a
distance to the fault place. At first these are non-
wide range of transients.
homogeneous feeders, distributed loads and many
The process of charging network capacity causes
branches. The impact of fault resistance, especially for
voltage disturbance in healthy phases. Because of voltage
single-phase faults in distribution system should be taken
drop of faulty phase to ground potential, a voltage of
into account. A load model is needed to predict load
neutral point of transformer in supply substation
behavior depending on changes in supply parameters
increases, which leads to increase in phase voltages of
(voltage and frequency).
healthy phases. Depending on fault resistance, earth
Loads distributed along distribution feeders can make
capacitance of healthy phases of all feeders in a network
significant errors during fault location. Although it is
have been charged through the inductance of transformer
tried to make a symmetry, during connection single-
to a voltage whose effective value is between phase to
phase loads at low voltage (0,4 kV), it cannot be said that
ground and phase to phase voltage. It is necessary to
loads in the distribution system are completely
analyze a transient fault that can be used for the purpose
symmetric. In the process of fault location is essential
of fault location. Typical fault transients can be classified
dynamic behavior of loads and here it is necessary to
as low-frequency and high-frequency transients. Low-
take into account the impact of the fault duration or the
frequency transients are related to charge transients and
impact of the time required to make the estimation of the
transient currents of compensation windings (in case of
distance to the fault [3]. In general, a fault time interval
compensated distribution networks). High-frequency
depends on several factors; value of a fault current,
transients include discharge transients of network
protection system in area where fault occurs, type of
capacity and traveling waves.
fault, etc. If fault location based on transient phenomena
Charge transients are relatively low frequencies, so
are used, it should be noted that electromagnetic
that the required level of the sampling frequency is a few
transients on distribution level are enough attenuated and
kHz. Travelling waves, in the case of overhead lines,
disappear for a few [ms].
spread around the speed of light. Even if a network is
Practically, electromagnetic transients never last
relatively long, it is necessary a considerable high
longer than a few cycles. Distribution feeders are
sampling frequency to identify and record traveling
comprised of laterals that have different electrical
waves. For overhead feeders, the minimum required
parameters so that appearance of discontinuities in the
sampling rate is in the order of one MHz. Charge
direction of the impedance is occurred, i.e., variations in
transients are low frequency, and in distribution
feeder impedance. It is caused by different types of
networks, have frequencies in the range of 100-800 Hz.
conductors, cross sections, geometry of stubs, different
The frequencies are near the upper limit for a faults
types of cables, switching from overhead to a cable
close the buses. In the case of method testing, transient
network, etc. For this reason, determination of a distance
spectrum of currents and voltages of faulty phase (or zero
to the fault point on the basis of estimated parameters
sequence current and voltage) was first observed in order
from supply substation to the fault point, such as
to determine the dominant frequencies [8]-[10].
reactance, inductance and impedance can be difficult. In
The spectrum of transient voltage and current was
such cases it is necessary to take data from a program for
extracted by using fast Fourier transform on signal of
network analysis where all discontinuity points are
voltage and current for the first cycle from the moment
identified and determined by values of short-circuit
of a fault inception. Fast Fourier transform (FFT) is a
currents.
complex linear transformation of a time domain to the
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
72
Š. Gruhonjić Ferhatbegović
Secondary side
frequency domain and thus provides an estimation of of transformer in TS x/10(20) kV
frequency components in input signal [11], [12]. Overhead feeder
Phase A
Ltr
With established charge and discharge frequencies for Lv
Cm
Vf
t
1
v0 t v0 t0 i0 d (1)
Cz t
0 Fig. 2. The equivalent scheme for the charge transient
frequency estimation
where is:
Cz – equivalent ground capacitance of a feeder, In the case of a fault on the buses 10(20) kV, it holds
I0 – zero current of a feeder that Lv=0, so that angular frequency of charge transient
For the estimation of charge transient frequency of has a slightly higher value:
network capacity, equivalent scheme with lumped
parameters, as indicated in Fig. 1, has been observed. 1
(3)
In the first approach the effective resistance of 3 Ltr C z Cm
conductors and transformers in the supply TS x/10(20)
kV were neglected. It was considered network with Eq. (3) is valid in the case where the inductance of
isolated neutral point and the direct earth fault, i.e. the transformer is large in comparison to the feeder
situation with zero fault resistance was assumed. inductance. In specific cases it is necessary to make a
During a fault in phase C, ground capacitance of comparison between transformer inductance and feeder
faulty phase can be neglected because it is short-circuited inductance whereas, when applying fault locator based
by fault itself. Phase to phase capacitances Cm of adjacent on charge transients, assign one particular frequency.
phases are parallel with their earth fault capacitances Cz, The assigned frequency need to be applicable for fault
as shown in the Fig. 1. The capacitance between phases in different places on a feeder. The initial amplitude of
that are not faulty (phases A and B in the analyzed case) the charge transient, without attenuation, can be
can be omitted because the ends are at the same potential calculated as:
with a transient viewpoint. The inductance of the
transformer substation in the supply TS x/10(20) kV is vk
Ltr. In accordance with the previously used labels the I vk Ceq (4)
Leq
equivalent diagram was shown in the Fig. 2.
According to this situation, the charge transient
frequency of phases that are not faulty can be calculated If a fault occurs at the moment of passing through the
from the following equation: phase voltage maximum, the amplitude of a charge
transient is given in the following equation:
1
(2) 2 Ceq
3 Ltr Lv C z Cm I Iz (5)
3 C z 1
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
73
Š. Gruhonjić Ferhatbegović
B1
In the case of networks with compensated neutral,
charge frequency transients is approximately the same as Fault
locator
in the case of networks with isolated neutral. The L1 L2 L3
resistance that compensation coil provides toward v(t) Supply TS x/10(20) kV Main sections:
passing of charge transients has great value. L1=9 km, L2=14 km, L3=8 km
B2
For example, if it is assumed charge transients Branches:
B1=11 km, B2=7 km
frequency of 400 Hz and a compensation coil inductance
of 5 H, then the inductor resistance toward passing of a Fig. 3. Example of 10 kV distribution feeder where computer
charge transients is 2000 Ω. Therefore, the path across simulations were made
charge transients are closed is approximately the same as
in the case of an isolated network. Also, this is shown by For this model the frequency spectrum of transients
computer simulations. In a network with low-resistance was observed. During testing, it was necessary to obtain
grounding, estimated value of frequency transients, an insight into the parameters that influence to the
according to an equivalent scheme, should correspond to spectrum of current and voltage. Therefore, a fault
the following relation: resistance, fault time, initial angle of faulty phase
voltage, treatment of neutral point, dynamic of loads and
location of a fault were changed. For example, with
1 1
Ruz j Ltr Lv modeled ground fault on the section L2, spectrum of
2 C z Cm current and voltage of faulty phase, obtained by Fourier
1 1 (6) transformation, was indicated in the Fig. 4. The spectrum
Ruz j Ltr Lv is calculated for the first cycle from the moment of a
2 C z Cm
fault occurrence.
j Ltr Lv 0 Selected signal: 7.5 cycles. FFT window (in red): 1 cycles
40
where is: 20
400
III. Transient Frequency Identification
by Computer Fault Simulation 200
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
74
Š. Gruhonjić Ferhatbegović
t1 t1 Vn
Ik3 (12)
v t dt Rx i t dt Lx i t1 i t0
(7)
t0 t0
3 R 2
x 2 L2x
t2 t2 The calculated value of current Ik3 is compared with
v t dt Rx i t dt Lx i t2 i t1 currents in short-circuit analysis in all nodes of the
(8)
considered feeder.
t1 t1
By identifying the nearest value of current and node
(discontinuity point) on a single line network diagram,
The previous integrals should be evaluated on the
the faulty area can be determined.
basis of samples of current and voltage. Integration of the
The difference between calculated fault current, which
Eqs. (7) and (8), for the equally spaced samples (for
correspond to estimated faulty path inductance and active
interval Δt), can be made by trapezoidal rule.
resistance, based on the charge transients and the nearest
For the samples, k + 1 and k + 2 can be made an
lower value of short circuit current, in an application for
estimation of faulty path resistance and inductance:
network analysis, was marked with ΔIk3:
vk 1 vk ik 2 ik 1 V1
I k 3 (13)
vk 2 vk 1 ik 1 ik 3 R2x 2 L2x
Rx (9)
ik 1 ik ik 2 ik 1
i i i i Distance from the nearest point of discontinuity to the
k 2 k 1 k 1 k
fault point is:
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
75
Š. Gruhonjić Ferhatbegović
30
V1
x (14) 20
3 I k 3 R'2x 2 L'2x
10
-30
the examination was done for the actual fault on the -0.5
network. The advantage of this approach is the ability to -1
use recorded transients from existing devices in the -1.5
network.
-2
It may be currents and voltages recorded by modern
-2.5
numerical protection and power quality analyzers that are 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
increasingly encountered in distribution networks. time
4000
fault from supply TS of 9 km, the results were shown in time (s)
200
seen that time interval with fault transients has duration
100
of one to, at most, two cycles.
0
-100
V.2. The Identification of Ground Fault Location
-200
in the Real Network
-300
In this case at the input of fault locator, signals of -400
current and voltage for actual fault in the real network
-500
were connected. 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
time
It was downloaded data measured by power quality
analyzer at 10 kV feeder [14]. The sampling frequency Fig. 5. Current and voltage of faulty phase and estimated charge
was 25.6 kHz. The registered values of current and transients. The network is isolated. The moment of the fault inception
voltage were given in the Fig. 7. The ground fault is t = 0 s. The initial angle of faulty phase voltage is θ = 240˚. Fault
occurred in phase C. resistance is Rf = 10 ohms
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
76
Š. Gruhonjić Ferhatbegović
150 4
x 10
1.5
vA
100
transients of current and voltage
1 vB
0 0
-50 -0.5
-100 -1
-1.5
-150 -0.1 0 0.1 0.2 0.3 0.4
0 0.005 0.01 0.015 0.02 0.025 0.03
time (s)
time (s)
17 300
iA
16 iB
200
iC
distance to the fault place (km)
13 0
12
-100
11
-200
10
9 -300
-0.1 0 0.1 0.2 0.3 0.4
8 time (s)
0 0.005 0.01 0.015 0.02 0.025 0.03
time (s) Fig. 7. Voltages and currents measured
in the supply TS 110/10 kV
Fig. 6. The extracted transients of current and voltage and the distance
to the fault point. The fault was simulated at distance of x = 9 km from 8000
the supply TS. The moment of the fault inception is t = 0 [s]. Fault vC
resistance is Rf = 50 Ω. Θ = 240°. The network is isolated 6000
4000
By examining the waveforms of voltage and current it
voltage vC(V)
50
highest values at a frequency of 450 Hz (charge
0
transients).
-50
It can be seen that charge transients of voltage reaches
the maximum value for x = 0,02, which corresponds to -100
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
77
Š. Gruhonjić Ferhatbegović
Y: 68.32
Y: 0.9921
0
[9] Š. Gruhonjić Ferhatbegović „Single phase fault location in
distribution network based on traveling waves“, Doctoral
-10
Dissertation, Faculty of Electrical Engineering and Computing,
May 2012. Zagreb, Croatia
[10] Ferhatbegovic, S.G., Marasic, A., Pavic, I., Single phase fault
-20
distance estimation in medium voltage distribution network based
on traveling waves, (2012) International Review of Electrical
-30
Engineering (IREE), 7 (1), pp. 3532-3541.
[11] Y.G. Paithankar, „Transmission Network Protection“, Theory and
-40
0 0.005 0.01 0.015 0.02 0.025 Practice, M.D. New York 1997.
time (s) [12] A. G. Phadke, J.S. Thorp, „Computer Relaying for Power
Systems “, John Wiley and sons, New York 1988.
Fig. 9. Voltage and current transients and estimated distance [13] Matlab – SimPowerSystems, User’s Guide, Hydro – Quebec,
to the fault point Mathworks, September 2009.
[14] Documentation of JP EP BiH: "Pilot projects for monitoring
power quality," Sarajevo 2011.
VI. Conclusion
It is possible to estimate fault location by using charge Authors’ information
transients. The estimation accuracy a lot depends on the Šeila Gruhonjić Ferhatbegović was born in
choice of transient frequencies. This means that it is 1969. in Bijeljina, Bosnia and Herzegovina. She
necessary to accurately extract the charge transients of received the B.Sc. and M.Sc. degrees from the
network capacity, which could constitute an aggravating Faculty of Electrical Engineering, University of
Tuzla, in 1994 and 2001, respectively. Doctoral
circumstance for the application of such conceived dissertation “Fault location of single phase fault
locators. Significantly attenuation of fault transients was in distribution network based on traveling
recorded for values of fault resistance greater than 200 waves” was defended at the Faculty of electrical
ohms. The limiting value of fault resistance, for reliably engineering and computing in Zagreb, in 2012.. Since 1994 she has
worked in JP EP BiH. She participated in the development of a number
fault location, ranges up to 250 ohms. of studies, reports and projects in JP EP BiH in the field of distribution
Therefore, the methods based on charge transients are systems, power analysis and power quality. As an author and co-author
not suitable for locating faults with high fault resistance she has published twenty seven papers at conferences and professional
in medium voltage distribution networks. The advantage meetings. At international conferences she has published seven papers.
of these methods is less required sampling frequency. In
these studies it was used sampling frequency of 5 kHz in
computer testing. In comparison with methods based on
traveling waves, it is possible to use data registered by
power quality analyzers, which is a significant
advantage.
References
[1] F. Božuta, „Automatski zaštitni uređaji u elektroenergetskim
postrojenjima“ (eng. “Automatic protection devices in power
plants“), Svjetlost, Sarajevo, 1989.
[2] Technical documentation and statistics of faults in the MV
network, taken from the Electric Power Industry of Bosnia and
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
78
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Keywords: Distribution Planning, Diversified Load Curves, Monte Carlo Simulation, Plug In
Electric Vehicles, Stochastic Modelling
Nomenclature I. Introduction
FAOC Free amount of chargers of batteries Power demand increment by Plug-in Electric Vehicles
LAOC Limited amount of chargers of batteries (PEV) penetration is an essential issue to consider
PEV Plug-in Electric Vehicles indistribution power system planning. This is because it
SOC State of Charge of an electric vehicle battery has several impacts in the power distribution systems
Ck Electric vehicle’s battery capacity (kWh) asthe power demand growth, variations in voltage
∆Ai Expected number of vehicles arriving at the profiles, increment in harmonic pollution, reduction of
parking lot over the total number of vehicles the lifetime of feeding components, power demand
which could park for period i response, and so on [1]-[20]. However, in some cities
∆Li Expected value of percentage of electric does not exist information about the power demand
vehicles leavings parking lot increment due to PEV penetration, which means that an
∆x Distance traveled by an electric vehicle in a estimation of it is needed. Several models of PEV power
journey demand have been made [1]-[6] as a starting point for
F(Ai) Cumulative distribution fuction of the impact assessment. In general, the aspects considered by
interval I of electric vehicles arrivals to the these models are: behavior of PEV drivers, curve of
parking lot charging of PEV batteries and energy consumed by
F(Li) Cumulative distribution fuction of the vehicle due to its usage.
interval I of electric vehicles leaving the The first aspect to consider is the behavior of PEV
parking lot drivers,which defines the time of connection or
Ndays Number of days of simulation disconnection of the Electric Vehicle (EV)to the grid.
nev Number of electric vehicles associated to a However, this is a stochastic variable with high
distribution transformer dispersion because it depends upon several variables as
Nev Maximum number of electric vehicles traffic conditions of the city, labor schedules, electricity
associated to a distribution transformer tariff politics, and so on. That is whyit is assumed in [1]
Oi Parking occupancy at hour i associated to a and quantified as an hourly percentage of the total
distribution transformer amount of vehicles in residential garages. Other authors
∆Oi Variation of the parking occupancy Oiat the [2] defines this behavior with a cumulative distribution
interval of time i function of arriving time since 17:00-24:00 hh:mm in the
P(t,nev) Power demand consumption at period t by day. Another option is to assume the time of connection
nev electric vehicles associated to a of vehicles based on empirical experience on the city in
distribution transformer study, as in Malaysia’s case [3].
rdk Rate of discharge of the battery (kWh/km) The second issue to consideris the curve of charging
of PEV batteries.
79
M. A. Ríos, C. A. Rincón
This is related to the type of charger used and contains According to that, this paper proposes to estimate the
information about the power required by batteries. power demand incrementrelated to the penetration of
One option to include that information is considering electric vehicles in urban markets based on a stochastic
it with a discretized model, with different linear relations model.
between time and energy for intervals of time[4].On the Section II presents the model as the considerations
other hand, [11]-[13] present PEV battery charger models taken in count and section III y IV presents the results of
based on its circuital model. this simulation for two cases: Limited amount of
Also, information about the amount of energy chargers (LAOC) and free amount of chargers (FAOC).
consumed from the battery should be included. This is Finally, conclusions of this work are presented.
related to the discharge rate of the battery and the
distance traveled by the vehicle. The first one is defined
on the data sheet of the EV battery, and the second one is II. Hourly Diversified Load Curves Model
related to traffic, distances in the city and the behavior of The model proposed is based on hourly diversified
vehicles. In some references, statistical data of the city in load curves associated with the power demand to the
study is included[6], but in other cases a distribution penetration of electric vehicles in urban markets. It is
function of distance traveled is assumed and the traffic is because these curves are tools typically used in power
ignored[4]. distribution planning[24].
With the estimation of the power consumption of the However, statistical data of EV usage are null in cities
EVssome impact assessments could be made. One where the utilization of EV is starting. So, it is necessary
studyis about the lifetime of feed components as the to replace the statistical data with a simulation-based
transformers are also impacted by the penetration of PEV projection model that is, obviously, based on the
[7]. This impact in the lifetime could be reduced by the characteristics of electric vehicle usage and the
installation of rooftop photovoltaics [8] as well as recharging process of the batteries.
coordinated charging of PEV by the utility [9]. Also, the model proposed applies to multifamily
The PEV penetration also impacts on electricity dwellings and individual residences fed by the same
demand, prices and generation dispatch. For instance, MV/LV transformer, where the residents own a specific
some simulations and assessments had been done as the number of EVs.
one presented in [10]. According to that, the determination of the diversified
Reference [14] considers the time of use electricity load curves requires punctual information of total hourly
prices as an additional aspect to be considered in the PEV power demand for different number of electric vehicles
demand estimation,because it makes changes in the (nev) owned by the residents in a specific area. With that,
behavior of PEV drivers. the solution of the optimization problem described in (1)
Furthermore, in [4] a discrete event simulation was returns the constants (C1, C2, B) which describe the
implemented in order to study the PEV penetration diversified load curves. This optimization problem
impacts in power distribution systems, for the specific should be solved for every hour(t) in a day:
case of commercial buildings. A similar analysis is
presented in [15], but this assessment was focused in 2
unbalanced voltages caused by uneven distribution of 1
Nev c t c t n
PEV penetration among the phases. Besides, EPRI has 1 2 ev
performed this kind of analysis with the purpose is to min
C1 t ,C2 t ,B t P ev
t,n
(1)
nev 1
identify, define, and calculate the impact to particular B t
utility distribution system architectures considering total nev
PEV penetration levels as well as localized concentration
[16]. Other related studies are presented in [17]-[20]. The optimization problem minimizes a quadratic error
According to the worksexposed above, the mitigation function between the hourly diversified load curve and
of the impact of PEV penetration plays an important role the average power demand by vehicle, and it has the
in power distribution planning, that is why several decision variables (C1, C2 and B).
studies in this area had been made. As an example, [21] Therefore, an estimation of the total power demand in
presents an algorithm to accommodate high penetration hour t for a numberof electric vehicles nev is required. For
of PEVs in distribution networks, which is based on this estimation, nev varies from 1 to a maximum number
allocating distributed generation units.This algorithmcan of electric vehicles Nev. Moreover, it is generated based
help the local distribution companies to better assess the on the Mote Carlo simulation as shown in Fig. 1.
PEV impacts on their systems and mitigate them. An The estimation process starts with the initialization of
additional point of view is the consideration of multi- the average hourly total power demand matrix (P),
terminal low voltage direct current for the improvement assigning a zero to every cell. This matrix contains
of PEV integration in an existing power distribution information about the aggregated power demand at hour t
network [22]. due to the presence of nevelectric vehicles, and it is an
Besides, a fuzzy coordination algorithm for charging average value of the realizations in Monte Carlo
PEV is proposed in [23]. simulation.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
80
M. A. Ríos, C. A. Rincón
Copyright © 201
2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10,, N. 1
81
M. A. Ríos, C. A. Rincón
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
82
M. A. Ríos, C. A. Rincón
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
83
M. A. Ríos, C. A. Rincón
In these cases the power demand curve is divided in III. Limited Amount of Chargers (LAOC)
two, the first part is assigned from the arriving time to
the end of the day, and the second part starts at the As it was presented above, one application of this
beginning of the same day and finishes when the SOC is stochastic model has as goal to estimate a load profile
1. due to EV charging stations in multifamily dwellings.
However, since the amount of free spaces for energy
meters is very limited (one free space in some cases)it is
recommendedto measurethe energy consumption of the
battery charger with only one meter associated with the
communal account of the multifamily dwelling.
Based on this, an additional recommendation is to
install charging stations in communal parking spaces,
which are also limited. So, for the case of multifamily
dwellings, the amount of chargers is limited; and, this
restriction should be included to determine if a vehicle
could charge when it arrives in the parking.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
84
M. A. Ríos, C. A. Rincón
TABLE V
BATTERY CHARACTERISTICS
Vehicle Type Modeled vehicle Capacity (kWh)
Compact Sedan Mitsubishi i-MiEV 16
Mid-Size Sedan Nissan LEAF 24
Mid-Size SUV BYDe6 61.4
Full-Size SUV BYDe6 61.4
III.2. Results
With the input data describedabove, diversified
demand curves were generated for each hour in a day. As Fig. 5. Hourly diversified demand curve (22:00-21:00)
an example, diversified demand curve for time interval
20:00-21:00 is presented in Fig. 5.
Stochastic results from the Monte Carlo simulation are
presented as circles for each amount of electric vehicles
between 1 and 100 and diversified demand curve as a
continuous line. It shows the expected behavior for this
type of curves, a high power average for low amount of
users (EV) and a tendency to keep constant with high
number of them.
This typical tendency is evident for some intervals of
time, including (20:00-21:00), and Fig. 6 presents its
behavior by number of EV. Moreover, those time
intervals are peak hours of demand, and interval (19:00-
20:00) is the highestduring the day, it reachesan average
of 0.623kW/EV for high amount of vehicles. Fig. 6. Hourly diversified demand curves– limited number of chargers
On the other time intervals, a constant behavior for
any number of vehicles is observed; Table VI
summarizes those constant values. IV. Free Amount of Chargers (FAOC)
As expected, time intervals of residential parking low
The second type of housing in study includes
occupancy represent low electric power demand, as is the
individual houses connected to the same MV/LV
case of intervals between 9 a.m. to 1 p.m.. On the other
transformer. In this case, it is assumed that every house
hand, intervals in the range 4-9 a.m. has low power
has the possibility to install at least EV charger per
demand because EV end to charge around 1-3 a.m.. It is
electric vehicle. So, the number of chargers is limited just
also observed that high demand intervals (7p.m.-10p.m.)
for the number of vehicles, and any vehicle could be
has the expected behavior of a diversified curve.
connected at any time.
TABLE VI The input data used in this case is the same as in the
PEAK POWER DEMAND FOR LAOC case of limited amount of chargerscase;but, the ratio
Time interval Ppeak/VE between the amount of chargers and vehicles is 1 (i.e. the
0:00-1:00 0,1332 number of vehicles is equal to the number chargers).
1:00-2:00 0,0988
2:00-3:00 0,0789
As in the other case, hourly diversified demand curves
3:00-4:00 0,0704 estimatedhad the expected behavior for some time
4:00-5:00 0,0622 intervals, and the other ones got constant values.Intervals
5:00-6:00 0,0549 withconstant values of power demand are presented in
6:00-7:00 0,0528
7:00-8:00 0,0525 Table VII, and the rest of them is presented in Fig. 7.
8:00-9:00 0,0492 It is important to note that for a free amount of
9:00-10:00 0,0440 chargers the electric power demand in intervals with
10:00-11:00 0,0355 most arriving events (7p.m.-10p.m) is higher than the
11:00-12:00 0,0317
12:00-13:00 0,0691 case of limited amount of chargers. This is presented
13:00-14:00 0,1395 because all of the vehicles which reach its parking space
14:00-15:00 0,1764 could connect to the grid; it was not limited by the
15:00-16:00 0,1700 vehicles already connected. Because the possibility of
16:00-17:00 0,2079
17:00-18:00 0,2745
free connection, vehicles have a higher average state of
18:00-19:00 0,3904 charger in stable conditions than the LAOC case. So,
23:00-24:00 0,1182 time of charging for batteries tends to be lower in FAOC
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
85
M. A. Ríos, C. A. Rincón
than in LAOC, and the power requirement is lower in the In addition, the model has not considered batteries’
latest interval in night (23:00-24:00) and first intervals managing system; that could be included adjusting the
during the day (00:00-03:00). LAOC and the FAOC models.
TABLE VII
PEAK POWER DEMAND FOR FAOC Acknowledgements
Time interval Ppeak/VE
1:00-2:00 0,0603 This work was supported in part by COLCIENCIAS,
2:00-3:00 0,0413 CODENSA, EMGESA Grant 0197-2012.
3:00-4:00 0,0318
4:00-5:00 0,0256
5:00-6:00 0,0203
6:00-7:00 0,0193
References
7:00-8:00 0,0219 [1] M.de Nigris, I.Gianinoni, S.Grillo, S.Massucco, F. Silvestro,
8:00-9:00 0,0227 Impact evaluation of plug-in electric vehicles (PEV) on electric
9:00-10:00 0,0232 distribution networks,14th International Conference on
11:00-12:00 0,0246 Harmonics and Quality of Power (ICHQP), Sept.26-29, 2010.
14:00-15:00 0,1664 [2] H. Wang, Q. Song, L. Zhang,F. Wen, J. Huang, Load
15:00-16:00 0,1551 characteristics of electric vehicles in charging and discharging
16:00-17:00 0,2005 states and impacts on distribution systems,International
18:00-19:00 0,4019 Conference onSustainable Power Generation and Supply
19:00-20:00 0,8140 (SUPERGEN 2012,) Sept. 8-9, 2012.
20:00-21:00 0,7277 [3] C.H. Tie,C.K. Gan, K.A. Ibrahim, The impact of electric vehicle
21:00-22:00 0,3391 charging on a residential low voltage distribution network in
22:00-23:00 0,1807 Malaysia, 2014 IEEE Innovative Smart Grid Technologies - Asia
(ISGT Asia), pp.272-277, May 20-23, 2014.
[4] M.A.Rios, N.M.Pena,G.A. Ramos, N. Achury, L.E. Muñoz,
Evaluation of the impact of Plug-in Electric Vehicles on the load
profile of commercial buildings: a stochastic model, 2014 IEEE
PES Transmission & Distribution Conference and Exposition -
Latin America, Sept. 10-13,2014.
[5] Malik, F.H., Lehtonen, M., Saarijärvi, E., Safdarian, A., A
feasibility study of fast charging infrastructure for EVs on
highways, (2014) International Review of Electrical Engineering
(IREE), 9 (2), pp. 341-350.
[6] Alharbi, Abdulelah Yousef, Impact of plug in electric vehicle
battery charging on a distribution system based on real-time
digital simulator, M.Sc Dissertation, Dept. Elect. Eng., The
University of Tennessee at Chattanooga, Chattanooga, TE, 2013.
[7] Q. Gong, S. Midlam-Mohler, V. Marano, G. Rizzoni, PEV
charging impact on residential distribution transformer life,2011
IEEE Energytech, May 25-26 , 2011.
[8] T.J.Geiles, S. Islam, Impact of PEV charging and rooftop PV
penetration on distribution transformer life,2013 IEEEPower and
Energy Society General Meeting (PES), July21-25, 2013.
[9] M.A.S.Masoum,P.S. Moses,S.Hajforoosh, Distribution
Fig. 7. Hourly diversified demand curve – free number of chargers transformer stress in smart grid with coordinated charging of
Plug-In Electric Vehicles,2012 IEEE PES Innovative Smart Grid
Technologies (ISGT), Jan. 16-20, 2012.
V. Conclusion [10] J. Villar, C.A. Diaz, J. Arnau, F.A. Campos, Impact of plug-in-
electric vehicles penetration on electricity demand, prices and
A stochastic model to estimate the power demand thermal generation dispatch,2012 9th International Conference on
increment due to PEV penetration in urban markets is the European Energy Market (EEM), May 10-12,2012.
[11] P.S.Priambodo, W.Purnomo, A. Subiantoro, A. Muis, F. Yusivar,
proposed. This estimation is based on information about Transformerless high voltage and controllable current battery
the behaviors of car drivers in the city to study, the charger for e-car,2013 Joint International Conference on Rural
electric properties of the battery chargers for different Information & Communication Technology and Electric-Vehicle
sizes of vehicles andthe energy consumption from the Technology (rICT&ICeV-T), Nov. 26-28, 2013.
[12] Hu, Y., Liu, H., Damage analysis of lithium iron phosphate
battery due to the vehicle usage. batteries for pure electric vehicles, (2013) International Review of
This estimation represents a tool for distribution Electrical Engineering (IREE), 8 (2), pp. 722-728.
planners to study the impacts of the power increase due [13] M.P.Puentes, G. Ramos, Level 2 bidirectional charger for small
to the connection of plug-in electric vehicle chargers in electric vehicles: Topologies comparison,2013 Workshop on
Power Electronics and Power Quality Applications (PEPQA),
residency markets in two different cases, multifamily July 6-7, 2013.
dwellings and individual housing. [14] A.S.Bin Humayd, K. Bhattacharya, Impact of PEV penetration on
The proposed model has been developed considering distribution system planning considering time-of-use electricity
only residential customers associated to a distribution prices,2014 IEEE PES General Meeting Conference &
Exposition, July27-31, 2014.
transformer. So, the model requires further work to [15] S.K.Bunga, A.H.Eltom, N.Sisworahardjo, Impact of Plug-in
include charging EV in commercial places associated to Electric Vehicle battery charging on a distribution system, 2014
distribution transformers. IEEE Industry Applications Society Annual Meeting, Oct.5-9,
2014.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
86
M. A. Ríos, C. A. Rincón
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
87
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – Modern planning of power distribution systems faces significant changes over the
last few decades due to the massive introduction of distributed generation units. In this paper the
authors discuss the inclusion of location constraints for optimal allocation of DG units. The goal
function will be the minimization of cumulative average daily active power losses. Four types of
DG units will be considered; a solar park, a wind farm, a power station that does not depend on
an intermittent primary energy source and a small hydro unit. Each DG unit and network load will
be modeled with its own characteristic average daily power production or consumption curve. The
network load will consist of residential and industrial consumers. The problem will be solved
using genetic algorithm and realized in Matlab programming environment. Copyright © 2015
Praise Worthy Prize S.r.l. - All rights reserved.
88
R. Prenc, N. Bogunović, A. Cuculić
The Croatian Grid Code for the electric power system III. Network Model and
from 2006 defined that on medium voltage (MV) Power Flow Solution
network only DG units with output power range between
500 kW and 10 MW can be connected, and in this paper The main difference between the connection of a load
the authors will present an algorithm for individual and a DG unit is that when connecting a DG unit the
allocation of four types of DG units. feeder input current will drop and in the same time some
The DG units considered will be a solar park, a wind of the network node voltages will simultaneously rise.
farm, a power station that does not depend on an As a result the network will change its operation from
intermittent primary energy source and a small hydro passive to an active network. Although the lines closer to
unit. the HV/MV Primary Substation will now be unloaded,
the lines near the DG units (usually having lesser cross
section because of passive network design) may become
II. The Definition of the Distributed overloaded. Also, the power flows may reverse and the
Generation Allocation Problem node voltages may increase to extreme values.
An example of a DG connection to an existing
The goal in this paper will be to minimize cumulative distribution network feeder is displayed in Fig. 1.
average daily active power losses for all the lines in the
radial network in discrete 24 hour intervals, as displayed
in the following equation:
24 lines 24 lines
plossi, j min 3 Ilossi, j
2
f min Ri (1)
j 1 i 1 j 1 i 1
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
89
R. Prenc, N. Bogunović, A. Cuculić
The equation that determines the three phase Unlike the crossover operator used in solving optimal
distribution line active power loss is: allocation of capacitor banks [15], the crossover point
used here is fixed and positioned in the middle of a
2
plossi 3 I lossi Ri chromosome. That means that a new offspring will have
the DG location from one parent and the DG size from
2
Pi cos i Qi sin i the other parent. The main reason why a fixed crossover
point is used lies in the integer coding of the
3U n chromosomes, which allows the distinct separation of the
Ci U n sin i DG location and size.
2 3 (6) Finally, the mutation used in this paper has two
3 2
Ri components, each one having its own predetermined
Qi cos i Pi sin i probability. The first type of mutation is the location
3U n mutation (with its predetermined probability ploc) and the
Ci U n cos i second type is the size mutation (with its predetermined
probability psize). If an individual chromosome undergoes
2 3 both mutations, it will simultaneously change its position
and size.
In the above equation Ci denotes parallel capacitance During the algorithm each chromosome will undergo
of line i. Pi and Qi represent active/reactive power that is a process called elimination load flow on three separate
injected in node n, Un is the line voltage of the nth node occasions. The first will occur while forming the initial
and φi represents phase angle between voltage and population, the second after the crossover operator and
current. The aforementioned equation will be used to finally the third after the mutation operator has been
evaluate every chromosome while performing the genetic performed. The elimination load flow has a purpose to
algorithm for DG allocation. examine voltage and thermal constraints of a network in
a worst case scenario where the individually observed
DG unit is online and operating with full peak power
IV. Genetic Algorithm while the network load is at its minimal level. If the
In this paper the authors will use a genetic algorithm elimination load flow is unsatisfactory during
to solve the DG allocation problem. Genetic algorithm initialization, the chromosome will be discarded and a
(GA) belongs to a class of heuristic optimization new random chromosome will take its place until the
methods that imitate the process of natural evolution, initial population reaches its predetermined size. If the
called evolutionary algorithms (EA). In this paper the elimination load flow is unsatisfactory during crossover,
authors have modified the genetic algorithm for DG the child chromosome will be discarded and a new
allocation problem. The coding of the individuals in random pair of parents will be chosen to create a new
population into chromosomes was done using integer child.
values, with every chromosome containing two parts. Finally, if the elimination load flow is unsatisfactory
The first part refers to the location of a DG unit and during mutation, the mutated child chromosome will be
the second part refers to the size of a DG unit, as discarded and the original child chromosome will again
displayed in Fig. 3. undergo the mutation process. It is important to note that
the elimination load flow is a single load flow analysis
and is not used for calculation of cumulative average
daily active power losses.
The cumulative average daily active power losses are
calculated during evaluation of each chromosome in the
current population. The input parameters for the
evaluation process are the average daily power
consumption curves for all the residential and industrial
customers. While evaluating the fitness function for each
Fig. 3. The coding of an individual solution into a chromosome chromosome, every chromosome must be decoded and
the average daily power production curves must be
Furthermore, the selection chosen for this problem formed for every DG unit. The curves will be formed
was a tournament selection, which involves running using the data provided by IPPs based on their
several "tournaments" among k individuals randomly preliminary measurements or power production
chosen from the current population. It must be noted that estimation.
the algorithm used in this paper has a built-in elitism, Finally, the decoded chromosome will undergo the
which means that the best individual in a generation is optimization load flow, which must be conducted 24
automatically stored and forwarded to next generation. times during an average day. The optimization load flow
The crossover operator the authors have chosen is a calculates the losses of an entire feeder in one hour of an
single-point crossover according to a predetermined average day in the year.
crossover probability (pc).
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
90
R. Prenc, N. Bogunović, A. Cuculić
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
91
R. Prenc, N. Bogunović, A. Cuculić
350
300
200
150
Node 15
100 Node 28
Node 7
50
every hour during the period of a single year. All the data 0,10
must be collected and processed to calculate irradiation,
0,00 Solar park
wind speed and water discharge values for an average
day of the year.
However, these values for each hour of an average day Time [h]
must be divided with maximal recorded irradiation, wind
speed and water discharge during the preliminary Fig. 9. The estimation of average daily active power production curve
measurement period. for a solar park in percentage values
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
92
R. Prenc, N. Bogunović, A. Cuculić
0,60
It is presumed that the data on ratio of measured water
0,50 discharge and maximal water discharge for an average
Paverage/Ppeak [%]
0,40
day in the year is displayed in Fig. 12.
0,30 0,90
0,80
0,20
0,70
Paverage/Ppeak [%]
0,10 Wind f arm 0,60
0,50
0,00
0,40
0,30
0,20 Small hydro
0,10
Time [h]
0,00
0,60
individually allocated solar park is at node 7 with peak
0,50 power of 7 MW, as displayed in Table III.
0,40 Also, the table contains the percentage reduction of
0,30 cumulative average daily active power losses compared
0,20 Power station to the base case distribution network without DG
0,10
0,00
connection.
The optimally allocated solar park reduces cumulative
average daily active power losses up to 40.8% compared
Time [h]
to the base case without DG connection.
In Fig. 13 the dynamic daily distribution of active
Fig. 11. The estimation of average daily active power production curve power losses is displayed for a base case without DG
for a power station in percentage values connection and after optimal allocation of a solar park.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
93
R. Prenc, N. Bogunović, A. Cuculić
TABLE III 50
40 15
Active power losses with an
allocated solar park
35 10
30 5
25 0
20 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time [h]
15
10 Fig. 14. The comparison of active power losses for each hour
before and after the wind farm allocation
5
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Also, the table contains the percentage reduction of
Time [h] cumulative average daily active power losses compared
to the base case distribution network without DG
Fig. 13. The comparison of active power losses for each hour connection.
before and after the solar park allocation
TABLE V
It can be seen that the losses become significantly THE ALLOCATION RESULTS FOR A POWER STATION
lower during the day when a solar park is activated, Optimal OPTIMAL Losses after power Reduction
especially around midday. position SIZE [KW] station allocation [kW] [%]
25 3,000 216.35 60.1
VI.2. The Allocation of a Wind Farm The optimally allocated power station reduces
The calculated optimal position and size of an cumulative average daily active power losses up to
individually allocated wind farm is at node 11 with peak 60.1% compared to the base case without DG
power of 3.750 MW, as displayed in Table IV. Also, the connection. In Fig. 15 the dynamic daily distribution of
table contains the percentage reduction of cumulative active power losses is displayed for a base case without
average daily active power losses compared to the base DG connection and after optimal allocation of a power
case distribution network without DG connection. station that does not depend on an intermittent primary
energy source.
TABLE IV It can be seen that the losses become significantly
THE ALLOCATION RESULTS FOR A WIND FARM lower during the most part of the day, because the output
Optimal Optimal size Losses after wind farm of a power station during the average day in the year is
Reduction [%]
position [kW] allocation [kW]
11 3,750 216.82 60.0
relatively constant.
However, during the night the losses actually increase
because the load is at its minimal value.
The optimally allocated wind farm reduces cumulative
average daily active power losses up to 60% compared to 50
lower during the most part of the day, because the output 25
of a wind farm during the average day in the year is
relatively flat. However, during the night the losses 20
10
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
94
R. Prenc, N. Bogunović, A. Cuculić
Although the location constraints limited the Since the lateral branch had a light load, the
connection of a power station to a lateral branch elimination load flow prevented the connection of a
(between nodes 25–30), it is obvious that the losses have hydro unit with bigger size. In this example it can be
decreased the most compared to the allocation results of observed how the location constraints have a great
other DG units. influence on network losses. For that reason the location
If the location constraints were removed and the constraints must not be neglected in the DG allocation
power station could connect on a main feeder branch algorithm.
(between nodes 1–17), the losses would be even more
reduced.
VII. Conclusion
VI.4. The Allocation of a Small Hydro Unit In this paper the authors have presented an algorithm
for the individual allocation of DG units based on
The calculated optimal position and size of an average daily power consumption and production curves.
individually allocated small hydro unit is at node 19 with The goal was the minimization of cumulative average
peak power of 2.250 MW, as displayed in the Table VI. daily active power losses. Using the proposed algorithm
Also, the table contains the percentage reduction of the obtained allocation results are much more detailed
cumulative average daily active power losses compared and precise, which in turn can have a great importance in
to the base case distribution network without DG avoiding unnecessary and often quite considerable
connection. connection costs. Also, the costs of the DSO will be
minimized, since the DSO is responsible for the
TABLE VI
generated losses in the distribution network, and must
THE ALLOCATION RESULTS FOR A SMALL HYDRO UNIT
Optimal Optimal size Losses after small Reduction
acquire energy on the electricity market to cover those
position [KW] hydro unit allocation [kW] [%] losses.
19 2,250 469.71 13.4 The next step in the authors' research is to integrate
energy storage in the distribution network (small
The optimally allocated small hydro unit reduces pumped-storage hydroelectricity, sodium-sulfur NaS
cumulative average daily active power losses up to batteries) and to incorporate more strict voltage
13.4% compared to the base case without DG constraints in the algorithm. In some Central European
connection. In Fig. 16 the dynamic daily distribution of countries a common practice is that the voltage variation
active power losses is displayed for a base case without of each feeder node must not be greater than ±3% after
DG connection and after optimal allocation of a small the connection of DG unit(s). The aforementioned topics
hydro unit. will be the subject of the next paper.
It can be seen that the losses become lower during the
most part of the day, because the output of a small hydro
unit during the average day in the year is relatively Appendix
constant. However, during the night the losses actually In Table A1 the electrical parameters of lines used in
increase because the load is at its minimal value. the network example are presented and in this paper one
The reason that the loss reduction was not so type of electric underground MV cables is used.
significant (only 13.4%) are the location constraints
which limited the connection of a small hydro unit to a TABLE A1
lateral branch (between nodes 19–21). DATA OF THE ELECTRICAL CABLES USED IN THE CASE STUDY
DISTRIBUTION NETWORK
50
Un R1 X1 B1 Inom
Type
45
Active power losses without (kV) (Ω/km) (Ω/km) (μS/km) (A)
previously connected DG units
NA2XS(F)2Y
20 0.206 0.115 82 345
40 3x(1x150) mm2
Active power losses [kW]
30
Table A2 contains the data regarding the maximum
measured amount of active and reactive power consumed
25
by every node in the network under study and the lengths
20 of electrical lines.
15
To calculate the minimum network load for the
elimination load flow it will be presumed that the data on
10
minimal amount of active and reactive power is known
5
for the entire feeder (Pfeeder_min=1,200 kW and
0 Qfeeder_min=500 kVAr). The data can be obtained via
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time [h] SCADA system (System Control and Data Acquisition)
which actively monitors the power on all the feeders of a
Fig. 16. The comparison of active power losses for each hour Primary Substation and sends the data to the DSO's
before and after the small hydro unit allocation
Control Centre.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
95
R. Prenc, N. Bogunović, A. Cuculić
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
96
R. Prenc, N. Bogunović, A. Cuculić
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
97
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
H. Laaksonen
Abstract – Short-circuit protection operation time delays in medium-voltage (MV) networks have
traditionally been dependent on fault-current magnitude or measured impedance with fixed time
delays or inverse time curves. However, MV feeder protection selectivity issues with low-voltage-
ride-through (LVRT) curves of distributed generation (DG) units and possibility of intended island
operation must be considered increasingly in the future as a part of the protection scheme.
Therefore, this paper proposes a future-proof, directional short-circuit protection scheme for
Smart Grids. The proposed scheme enables the definition of LVRT curve compatible protection
operation time delays during normal, grid connected operation for protection relays in MV
distribution network having multiple protection zones without use of high-speed communication
based interlockings/blockings between relays. Copyright © 2015 Praise Worthy Prize S.r.l. - All
rights reserved.
98
H. Laaksonen
unit connection or disconnection. In such case where between LVRT curve of DG units (defined by grid
both inverter-based and synchronous generator based DG codes) and required time differences between CB2 and
can be simultaneously connected to the same MV feeder, CB3 in forward direction may be hard to achieve.
operation time delay curve of the proposed protection This naturally depends from the number of
scheme must be based on the most stringent DG unit consecutive protection zones and the allowed time
LVRT curve. difference between operation time delays of CB2 and
Effect of other recent and forthcoming grid code CB3.
requirements, stated for example in [9]–[12], has been
also studied in this paper. Reason for this was to confirm
that other requirements, such as converter based DG
units’ voltage support by reactive current injection, active
power/frequency (P/f)-or reactive power/voltage (Q/U) -
control of DG units will not affect to the proposed
protection scheme negatively. Previously in [13] some of
these new grid code requirements have been briefly
summarized. Also in [14] the risk of islanding situations
stabilization due to P/f- and Q/U-control requirements
has been studied.
In following, first possible time selectivity issues with
DG unit LVRT curves and traditional protections
schemes are shortly highlighted when MV feeder
consists from several consecutive protection zones. After
that, the proposed grid code compatible protection
scheme is presented. Finally, example simulation results
are presented followed by discussion and conclusions.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
99
H. Laaksonen
III. Proposed Grid Code Compatible infeed of multiple DG units in that zone (both MV and
Protection Scheme LV network connected DG units) is taken into account
and discrimination between 3-phase short-circuit faults in
In Section II (Fig. 1) it was presented how time LV busbars or LV feeders at MV/LV substations could
selectivity with DG unit LVRT curves may be be still done.
challenging to achieve, if fixed time delays are used, MV
feeder consists of many consecutive protection zones and Reverse (rev) Direction
high-speed communication utilization is not possible. Short-circuit protection fault detection (pick-up/start)
In Section I some of the main features of the proposed in reverse (rev) direction is based on exceeding
grid code compatible protection scheme were shortly undervoltage pick-up/start value (Fig. 2). Overcurrent-
stated. In Fig. 2 and in the following a more detailed based pick-up is not applicable, because fault current
description about this scheme will be given. levels of inverter-based DG units can be quite low. In
addition, final detection in reverse direction requires
correct current direction determination. In comparison to
detection in forward direction, the detection in reverse
direction is by default not allowed to give operate signal
to corresponding circuit-breaker (Fig. 2).
However, depending from the islanding possibility
(i.e. power balance through CB before fault) and
allowance to utilize intended island operation, operate
signal can be allowed (Fig. 2). In other words this means
that transition to intentional island operation is only
possible if active and reactive power unbalance (Punb and
Qunb) through connection point CB is small enough (or
enough, rapidly controllable active and reactive power
units exist in the possible island) before protection
start/operation in reverse direction. If this is not the case
transition to island operation should be not allowed.
During meshed or ring operation of MV feeders pick-
up/start criteria is similar in forward and reverse
directions.
Direction Determination
Based on PSCAD simulations, in addition to
traditional methods for fault current direction
determination, Upos_seq_angle- Ipos_seq_angle(Upos_seq_angleand
Ipos_seq_angleare voltage and current positive sequence
angles respectively) based direction determination is
possible during both 3- and 2-phase short-circuit faults as
well as during 2-phase earth-faults. Also U0_angle - I0_angle
Fig. 2. Proposed directional short-circuit protection scheme during
normal, grid-connected, radial operation of MV network (U0_angle and I0_angle are voltage and current zero sequence
angles respectively) based angle determination could be
possible in 2-phase-earth-faults in MV networks with
III.1. Directional Protection Pick-up/Start and isolated neutral treatment.
Direction Determination
Forward (fwd) Direction III.2. Voltage Dependent Time Delay
With short-circuit protection in distribution networks
having inverter or synchronous generator-based DG Time Delays with Even Time Steps (tstep) – Low Number
units, the detection of fault i.e. pick-up/start in forward of Protection Zones
(fwd) direction is based on exceeding over-current pick- After fault detection in forward or reverse direction
up/start value AND/OR undervoltage pick-up/start value (Fig. 2) voltage dependent time delay of the proposed
(Fig. 2). Also 2-phase earth-faults are included in protection scheme is determined. The main idea is to use
proposed short-circuit protection scheme (Fig. 2). In similar LVRT curves for calculation of MV feeder IED
addition, for final detection in forward direction also protection time delays (i.e. voltage dependent time
correct current direction determination is required. delays) than are required from the DG units by the grid
Also impedance or admittance based pick-up/start codes. In this way it is easier to achieve selectivity with
value could be possible with this scheme and then DG unit LVRT curves. Time delay of the proposed
impedance start value should be at least 105-110% of scheme is dependent from voltage dip magnitude i.e.
protected zone (MV line length) so that also possible distance from the fault so that the protection operation
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
100
H. Laaksonen
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
101
H. Laaksonen
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
102
H. Laaksonen
Case 1 – Radially Operated MV Feeders Case 3 (Radial) – Effect of Change from Dual-Breaker to
Single-Breaker Scheme
Fig. 6. Case 2 with radially operated MV feeder, Fig. 8. Case 1 with ring operated MV feeders
effect of topology change (See Figs. 3, 4 and 5) (See Figs. 3, 4 and 5)
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
103
H. Laaksonen
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
104
H. Laaksonen
Operation time in reverse direction trev is in brackets in control with 40 ms time delay wouldn’t ever have an
Table III, and also on other following Tables, if islanding effect on the performance of the proposed scheme.
is not possible or feasible from point of view of power On the other hand, many of these ‘equal or less than
unbalance (Fig. 2). 40 ms frequency transients’ can be a result from nearby
fault. Therefore, another possibility to prevent
TABLE III unnecessary operation of P/f -droop control with very
OPERATION TIME DELAYS TFWDAND TREV FOR DIFFERENT CBS IN CASE 1 short operation time delay (less than 40 ms) could be
WITH INVERTER-BASED DG UNITS 1 AND 2 (FIG. 9 AND TABLE I AND
EQ. (1) AND (2)) simultaneous use of additional clause like Uph-ph_min<
Operation time delays tfwd and trev for different CBs 0.2*Upos_pu, where Uph-ph_min is the minimum phase-to-
in Fig. 9 phase voltage and Upos_puis the positive sequence voltage
A1 A2a A2b B1 B2 in per units. This additional clause would then prevent
tfwd /trev tfwd /trev tfwd tfwd /trev tfwd /trev unnecessary operation of P/f -droop control of LV
Fault
[ms] [ms] [ms] [ms] [ms]
F1 o)
95X) o)
(173)X) - o)
95X) o)
(173)X) network connected DG units during nearby 3-phase
F2 o)
95X) o)
(173)X) - o)
95X) o)
(173)X) short-circuits. In following this kind of under-voltage
o)
F4 671 173X) - o)
214 o)
(571) blocking has been used to prevent P/f -droop control
o) o)
F5 945 107 - 283 (805) operation in large voltage dips.
F6 947 107*) 35*) o)
285 o)
(818)
F7 2013 1025 582 o)
1090 o)
(1820) In Table IV, the operation time delays in forward (tfwd)
F8 o)
95X) o)
(173)X) - 178 o)
173X) and reverse (trev) direction for different CBs are presented
o) o)
F9 1339 (2011) - 2211 768 during different faults (Table I and Fig. 9) from case 1
*)
Here in dual-breaker scheme also blockings between A2a and A2b with inverter-based DG units 1 and 2 (Fig. 9) when the
could be beneficial to ensure selectivity, X) DG unit 1 (Fig. 9) could get effect of P/f -droop control of LV network connected DG
disconnected and transition to island operation could fail due to
ENTSO-E NC RfG[9]LVRT requirements, o) Small power unbalance units with 0 ms operation time delay and under-voltage
(Punb, Qunb) => Islanding allowed i.e. operation in reverse direction if blocking is studied.
enough rapidly controllable DG, energy storage or controllable loads
connected to intended island part of the network (largest possible island TABLE IV
could be prioritized) OPERATION TIME DELAYS TFWDAND TREV FOR DIFFERENT CBS IN CASE 1
(P/F -CONTROL) WITH INVERTER-BASED DG UNITS 1 AND 2 (FIG. 9 AND
TABLE I AND EQ. (1) AND (2))
IV.2. Case 1 (Radial) – Effect of P/f -droop Control Operation time delays tfwd and trev for different CBs
of LV Network Connected DG Units in Fig. 9
A1 A2a A2b B1 B2
In this Section IV.2 the effect of P/f -droop control of tfwd /trev tfwd /trev tfwd tfwd /trev tfwd /trev
LV network connected DG units is presented. In the Fault
[ms] [ms] [ms] [ms] [ms]
future, this P/f -droop control will be required by grid F1 o)
95 X) o)
(173)X) - o)
95 X) o)
(173)X)
o)
codes like [9] from all DG units. In the simulations the F2 95 X) o)
(173)X) - o)
95 X) o)
(173)X)
o)
effect of P/f -droop control of LV network connected DG F4 661 173 X) - o)
209 o)
(570)
o) o)
F5 958 107 - 287 (815)
units during over-frequency situations was simulated F6 926 107 *)
35 *) o)
280 o)
(796)
between 50.3 and 51.5 Hz by linearly increasing the F7 1989 1012 571 o)
1061 o)
(1795)
o)
loading in part of the MV/LV transformers. In the F8 95X) o)
(173)X) - 178 o)
173 X)
o) o)
simulations the initial assumption was that the intentional F9 1356 (2022) - 2211 768
*) X)
time delay of P/f -droop control of LV network , and o) See definitions from Table III
connected DG unit is 0 ms as also proposed in [10].
The reason for this was the fact that in theory with 0 When comparing the protection operation time delays
ms time delay the P/f -droop control might have an effect tfwd and trev between Table III and IV, it can be seen that
on the performance of the proposed protection scheme, they are quite close to each other. Some deviations exist
because the time delays are calculated 30 ms after pick- in faults F6 and F7 (Fig. 9) in trev and tfwdof CBs located
up. Otherwise, the simulated cases were similar with further from the fault (i.e. not closest to the fault).
Section IV.1.More details about some of these new grid
code requirements, like P/f -droop control, can be found IV.3. Case 1 (Radial) – Effect of Reactive Current
from [13]. Injection of MV Network Connected Inverter-
Also in this case with inverter-based DG units the based DG Units
simulations are done with 0 msP/f -droop control
operation time delay as proposed in [10]. However, in In following the effect of voltage support by reactive
[10] it has been stated that “the frequency protection current injection during voltage dips / faults with
shall be insensitive to frequency transients with duration inverter-based MV network connected DG units is
equal or less than 40 ms”. Therefore, from this point of presented. Simultaneously it has been assumed that P/f-
view one could conclude that also minimum reasonable droop control, studied in previous Section IV.2, is used
operation time delay for P/f -droop control could be 40 with 400 ms time delay. In this way it is also easier to see
ms. Because time delays in the proposed protection the only the effect of reactive current injection with
scheme are calculated 30 ms after pick-up, P/f -droop minimum operation time delay more clearly. This
reactive current injection -control is already required by
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
105
H. Laaksonen
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
106
H. Laaksonen
In this case it would mean that operation time delay of The proposed LVRT curve / network code compatible
A2b would increase (Table III) due to 0.1 pu increase in scheme enables the definition of protection operation
Uph_ph_min_puafter opening of A3 after F7 fault if large time delays during normal, grid connected operation for
enough (e.g. 0.15 pu) limit for change in Uph_ph_min_pu is intelligent electronic devices (IEDs) in MV distribution
not used. network having multiple protection zones without the
need for high-speed communication. The suggested new
TABLE VI
scheme is applicable for MV networks with different
OPERATION TIME DELAYS TFWDFOR DIFFERENT CBS IN RING
OPERATION OF CASE 1 WITH INVERTER-BASED DG UNITS 1 topologies (radial or ring). The proposed scheme takes
AND 2 (FIG. 10 AND TABLE I AND EQ. (1) AND (2)) also into account the possibility of transition to intended
Operation time delays tfwd for different CBs in Fig. 10 island operation. The proposed scheme is generally
A1 A2a A2b A3 B3 applicable with different LVRT curves defined in various
Fault
tfwd tfwd tfwd tfwd tfwd grid codes. Effect of other recent and forthcoming grid
[ms] [ms] [ms] [ms] [ms]
F4 X) 760 - - - - code requirements was also studied in this paper.
F5 X) 1046 178 - - - Reason for this was to confirm that other
F6 X) 1046 178*) 142*) - - requirements, such as converter based DG units’ voltage
F7 o) 1798 1138 933 - - support during voltage dips by reactive current injection
F8 Y) -
- - - -
F9 o) 1798 1138 933 107*) 71*) (Iq_inj), active power/frequency (P/f)- or reactive
*)
During ring operation also blockings needed between A2a and A2b to power/voltage (Q/U) -control of DG units will not affect
maintain selectivity, X) DG unit 1, o) DG unit 2 or Y) DG units 1 and 2 to the proposed protection scheme negatively. However,
(Fig. 10) could get disconnected and transition to island operation could no significant effects were found in the simulations. As
fail due to ENTSO-E NC RfG [9] LVRT requirements one can see from the simulation results during radially
TABLE VII operated MV feeders (Fig. 9) in Tables II–V, the
OPERATION TIME DELAYS TREV FOR DIFFERENT CBS IN RING operation time delays in forward tfwd and reverse trev
OPERATION OF CASE 1 WITH INVERTER-BASED DG UNITS 1 AND 2 directions based on Fig. 3 and 4 may become sufficiently
(FIG. 10 AND TABLE I AND EQ. (1) AND (2)) long. Especially when the faults are at the end of the
Operation time delays trev for different CBs in Fig. 10
radial MV feeders like F7 and F9 faults in Fig. 9.
A2b A3 B1 B2 B3
Fault However, by using stricter tmin (and possibly also tmax)
trev trev trev trev trev
[ms] [ms] [ms] [ms] [ms] operation time delay curves than presented in Figs. 3 and
F4 X) Z)
71*) 197 *) 760 507 260*) Fig. 4 the operation time delays in this kind of faults at
F5 X) 71 231*) 1046 667 305*) the end of certain protection zone could be cleared much
F6 X) - 239*) 1046 675 315*)
F7 o) - 107*) 1798 881 143*)
faster. In Fig. 11 an example about choosing stricter tmin
F8 Y) - - 214 - - curve with inverter-based DG units than in Fig. 3(a)) is
F9 o) - - 1798 881 - presented.
*) X) o)
, , and Y)See definitions from Table VI,
Z)
35 mstrevin A2a*)
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
107
H. Laaksonen
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
108
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – The analysis of voltage instability in electric power system is very crucial in order to
maintain the equilibrium of the system. This paper presents the analysis of voltage instability of
electric power system by using power-voltage (PV) curve and reactive power-voltage (QV) curve.
This research focuses on the voltage instability analysis using PV and QV curves for radial-type
and mesh-type power networks.
For radial-type power network, the plotting of PV and QV curves will be done by using Graphical
User Interface (GUI) based on MATLAB application software. In the other hand, the power flow
analysis for mesh-type power network will be done by using Power World Simulator and the
plotting of PV and QV curve will be done by using Microsoft Excel. Copyright © 2015 Praise
Worthy Prize S.r.l. - All rights reserved.
Nomenclature I. Introduction
C Capacitance Voltage stability can be defined as the ability of a
GUI Graphical user interface power system to keep the voltage at all busses in the
IS Sending-end current system remain steady right after the system is being
IR Receiving-end current subjected to a disturbance. Voltage instability in the other
L Inductance hand is the opposite of voltage stability. Voltage
P Real power instability occurs due to the failure of the power system
PR Receiving-end real power to supply ample power to cover the increased demand of
PRmax Maximum receiving-end real power load. In addition, the characteristics of the load can also
PV Power-voltage contribute to voltage instability [1]-[3].
Q Reactive power Basically, there are three characteristics of loads.
QR Receiving-end reactive power Those characteristics are constant power load, constant
QRmax Maximum receiving-end reactive power current load and constant impedance load [3], [4].
QV Reactive-power voltage The load characteristics are depicted in the form of
R Resistance graph in Fig. 1 [4]. From Figure 1, it can be seen that the
SR Receiving-end apparent power power supplied to constant power load is constant even
SRmax Maximum receiving-end apparent power though the voltage of the load fluctuates. If the voltage
Vgap Voltage gap drops, more current will be drawn in order to maintain
VR Receiving-end voltage the amount of power. Most of industrial electrical motors
Vrated Rated voltage are of constant power load.
VRcri Critical receiving end voltage Fig. 1 also shows that the power consumed by
Vreg Voltage regulation constant current load is proportional to voltage. There are
VRFL Full-load receiving end voltage only a few electrical devices that are of constant current
VRNL No-load receiving end voltage load such as video tape recorders [3]. Lastly, constant
VS Sending end voltage impedance load uses power in proportional to voltage
Y Admittance squared. In other words, electrical devices that are of
Z Series impedance constant impedance load require less power when the
δcr Delta critical voltage is low. Incandescent lamps are one of many
δ'cri Prime critical angle examples of this type of load [3].
δ’ Delta prime Furthermore, Fig. 1 also depicts that the allowable
ϕ Power factor angle system voltage range is within ± 5% of rated voltage, i.e.,
ϕ' Phi Prime 1.0 per unit. This research focuses on constant power
load characteristics since this type of load is favorable to
voltage instability.
109
Marizan Sulaiman, Ahmad Fateh Mohamad Nor, Noor Ropidah Bujal
II.1. Example of Transmission Line ii. Using the power factor angle, calculate the phi prime,
ϕ' using;
Transmission line has been chosen for this study. The
line parameters that are related to transmission system ' 180 (4)
are Resistance (R), Inductance (L) and Capacitance (C).
Fig. 2 [14] shows the equivalent circuit of the
transmission line. From these R, L and C values, the iii. Determine the maximum receiving-end apparent
value of ABCD parameters can be obtained by using Eq. power, SRmax:
(1) [10], [14]:
Vrated 2
ZY S Rmax (5)
ZY '
A D 1 ; B Z ; C Y 1 (1) 4 AB sin 2
2 4 2
In matrix form, the generalized constant ABCD vi. SRmax is used to determine the maximum receiving-
parameters are: end power, PRmax :
VS A B VR
I C D I R
(2) PR max S R max cos (6)
S
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
110
Marizan Sulaiman, Ahmad Fateh Mohamad Nor, Noor Ropidah Bujal
v. SRmax is also used to determine the maximum C. Determine sending-end and receiving-end current,
receiving-end reactive power, QRmax : voltage and power, voltage regulation, voltage gap
and line current.
QR max S R max sin (7) i. The first step is to calculate the receiving-end current,
IR :
vi. Then, determine the receiving-end voltage at a critical PR max
point, VRcri : IR (17)
3 Vrated cos
Vrated
VRcri 3 phase (8) ii. To calculate the sending-end voltage the following
'
2 A sin calculation has been done:
2
VRcri 3 phase ZY
A D 1 ; B Z
VRcri 1 phase (9) 2
3 (18)
ZY
C 1 ; Y G jB
vii. The last step is to calculate prime critical power 2
angle, δ’cri :
' iii. Substitute Eq. (18) into the matrix Eq. (19) to
'cri 90 (10) determine sending-end voltage (Vs) and sending-end
2
current (Is) for medium-length transmission line.
and the critical power angle, δcri can be obtained by
using: ZY
cri 'cri (11) 1 2 Z V 0
Vs R1 ph
Is (19)
Y 1 ZY 1 ZY I R
B. Determine the other points. The other points are
4 2
calculated as;
i. Calculate the delta prime, '1 and '2 :
iv. Calculate the sending-end power, Ps:
'1 ' cr1 ; '2 ' cr2 (12)
Ps 3VsL-L Is cos Vs Is (20)
ii. Then calculate the receiving-end apparent power, SR1
and SR2: v. Receiving-end power, PR can be obtained by using:
iv. Calculate the receiving-end reactive power, QR1 and Vrated Vcri
QR2 : Vgap 100% (23)
Vrated
QR1 S R1 sin ; QR 2 S R 2 sin (15)
v. The last step is determining the receiving-end II.3. Graphical User Interface (GUI)
voltage, VR1 and VR2 :
The GUI for analyzing voltage instability for radial-
type electrical power system developed by [10] has been
Vrated sin ' '1 improved. In the latter version, the users only need to
VR1
A sin ' enter the values of R, XL and Y instead of ABCD
(16)
Vrated sin ' '2 parameters. The GUI based PV and QV calculations are
VR 2 depicted in Fig. 3.
A sin '
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
111
Marizan Sulaiman, Ahmad Fateh Mohamad Nor, Noor Ropidah Bujal
TABLE III
TRANSMISSION LINE DATA
From To Limit
R X B B*2
Bus Bus (MVA)
7 8 .0085 .072 .0745 .149 150
9 8 .0119 .1008 .1045 .209 150
7 5 .032 .161 .153 .306 150
9 6 .039 .17 .179 .358 150
5 4 .01 .085 .088 .176 150
6 4 .017 .092 .079 .158 150
TABLE IV
LOAD DATA
Fig. 3. GUI of voltage instability analysis Bus P (MW) Q (MVAR)
for radial-type electrical power system 5 125 50
6 90 30
8 100 35
III. Mesh-Type Electrical Power System
For mesh-type electrical power system, the voltage The methodological steps of plotting the PV and QV
instability analysis process involves a series of power curve for the 9-bus electric power system are listed as
flow solutions. The value of real power (MW) will be follows [3], [15], [18]:
increased constantly for every series of power flow. i. Select a bus. The PV and QV curve will be plotted at
The value of voltages of the bus will change as a this selected bus.
result of the series power flow [15]. Fig. 4 shows the ii. Run the power flow program by using Power World
mesh-type power network that consists of 9-bus power Simulator.
system constructed in Power World Simulator [16], [17]. iii. Increase the value of real power of load (P) by 0.1 per
The data for bus the 9-bus power system as shown in unit in order to plot PV curve. Similarly, increase the
Fig. 4 are presented in Table I until Table IV [16], [17]. value of reactive power of load (Q) by 0.1 per unit in
order to plot QV curve. Then run the power flow
TABLE I program again. A new voltage value at the selected
GENERATOR DATA bus will be obtained. Record the values of P or Q and
Rated Power Rated P Rated Q voltages.
BUS kV
MVA Factor (MW) (MVAR)
1 247.5 16.5 1.00 247.5 0 iv. Repeat step (iii) until the value of desired P or Q
2 192.0 18.0 0.85 163.2 101.1 cannot be delivered by the system.
3 128.0 13.8 0.85 108.8 67.4 v. Repeat step (iii) again for another several times to
obtain the lower point of the PV and QV curve.
TABLE II vi. Finally, use all the recorded value of P or Q obtained
TRANSFORMER DATA
Limit
in step (iii) – (v) and plot it against the voltage of the
From Bus To Bus R X Tap selected bus by using Microsoft Excel.
(MVA)
4 1 0 0.0576 1.00 300 The above steps will be applied on to Bus 5, Bus 6
7 2 0 0.0625 1.00 300 and Bus 8 of the 9-bus power system.
9 3 0 0.0586 1.00 300
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
112
Marizan Sulaiman, Ahmad Fateh Mohamad Nor, Noor Ropidah Bujal
IV. Results and Discussion Fig. 7 shows that the critical voltage points for both
PV and QV curves are 0.7032 per unit and 0.5301 per
IV.1. Results for Radial-Type Electrical Power System unit, respectively. The system will reach critical voltage
Case 1: Results for lagging power factor (18.2°). The point if the power load is increased until 515 MW or if
PV and QV curve is shown in Figure 5. the reactive power load is increased up to 310 MVAr.
Figure 5 shows the PV and QV curve at 0.95 lagging The green lines represent the allowable range of system
power factor which is a good power factor. Real power voltage within ± 5% of rated voltage, that is to say ± 5%
and reactive power are being represented by the black voltage regulation. The system voltage will operates
curve and red curve, respectively. The values of real and within ± 5% range if the value of real power (P) and
reactive power are positive because the loads that operate reactive power (Q) of load do not exceed 305 MW and
at lagging power factor absorb both real and reactive 100 MVAr, respectively.
power. The voltage regulation is 24.3 % and the value of Similarly, Fig. 8 shows critical voltage points for both
voltage gap is 38.5%. The rated voltage (blue line) is PV and QV curves generated at Bus 6 at 0.7394 per unit
operated far from the critical voltage. This is a very good and 0.5259 per unit, respectively. These points can be
situation in order to prevent voltage instability. Voltage reached if the power or reactive power load demand are
instability will occur if the rated voltage operates at or increased up to 470 MW or 299 MVAr, respectively. In
near the critical voltage. order to make sure that the system voltage always
operates within the range of ± 5% of rated voltage, both
Case 2: Results for leading power factor (-18.2°). The of the real power or reactive power of load demand must
PV and QV curve is shown in Figure 6. not exceed 290 MW and 90 MVAr, respectively.
Fig. 6 shows the PV and QV curve at 0.95 leading Finally, Fig. 9 shows voltage instability can occur at
power factor. It can be seen from this curve that the value Bus 8 if the real power or reactive power of load demand
of real power is positive but the value of reactive power rises until 510 MW (corresponding to V = 0.8381 per
is negative. This is due to the loads that operate at unit) and 115 MVAr (corresponding to V = 0.6812 per
leading power factor absorb real power but generate unit), respectively. The system voltage will operate
reactive power. The rated voltage (blue line) also below the ± 5% allowable range if the value of real
operates far from the critical voltage. The voltage power of load demand exceeds 370 MW or the value of
regulation is 13.9 % and the value of voltage gap is reactive power of load demand exceeds 115 MVAr.
20.9%. Lower voltage regulation is due to the voltage It has been shown in Fig. 7, Fig. 8 and Fig. 9 that the
rise at the receiving-end (current injection) and lower increase of real power (P) or reactive power (Q) of load
voltage gap is susceptible to voltage instability compare will force the bus voltages to decrease. Table V shows
to the lagging load conditions. the details of Fig. 7, Fig. 8 and Fig. 9 in terms of the
incremental values of P and Q to reach critical voltage,
ΔP and ΔQ. On the other hand, the incremental values of
IV.2. Results for Mesh-Type Electrical Power System P and Q to reach the minimum limit of allowable system
The generated PV and QV curves for Bus 5, Bus 6 and voltage range (that is 0.95 per unit), ΔP1 and ΔQ1.
Bus 9 are shown in Fig. 7, Fig. 8 and Fig. 9, respectivel.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
113
Marizan Sulaiman, Ahmad Fateh Mohamad Nor, Noor Ropidah Bujal
V. Conclusion
Voltage instability analysis is an important parameter
for monitoring the bus voltage in the electrical power
system. The bus voltage must not operate close to the
critical voltage point at nose point of PV and QV curves.
For lagging loads condition of radial-type power
system, both real and reactive powers of load are
positive. This indicates the load absorbs both P and Q.
On the contrary, in leading load condition, the real power
is positive but the reactive power is negative. This means
that during leading load condition, the load absorbs P but
Fig. 8. PV and QV curves at Bus 6 generates Q.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
114
Marizan Sulaiman, Ahmad Fateh Mohamad Nor, Noor Ropidah Bujal
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
115
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Sadik Farhat, Rachid Alaoui, Abdelilah Kahaji, Lahoussine Bouhouch, Ahmed Ihlal
Abstract – The maximum power point (MPP) of a photovoltaic (PV) system varies with
temperature and solar irradiation. The implementation a technical control of the PV power
converter systems to search and make this systems to operate at maximum output power is
required. The realization of the aforementioned converter and digital implementation of this
particular control are presented with experimental results. Hill climbing techniques, which allows
tracking the MPP in real time, is implemented.
This technique involves in the implementation and comparison of the efficiency of both methods:
"Perturb and Observe" (P&O) and the Incremental Conductance (InC). The proposed techniques
are validated using the experimental data of 40Wp PV panel (PVP), controlled by a boost
converter. Following the comparison of both chosen techniques, output power efficiency and
dynamic response are discussed. We also present the advantages and drawbacks of two techniques
by using the performed measurements results. Copyright © 2015 Praise Worthy Prize S.r.l. - All
rights reserved.
116
S. Farhat, R. Alaoui, A. Kahaji, L. Bouhouch, A. Ihlal
TABLE I Iopt
Ipv (A)
20
We use a test bench that enables us the evaluation of
the efficiency and limitations of two proposed techniques
(P&O and InC). For the validation, we implement and we 0
test their algorithms that appropriate on a PV system 0 8 Vpv (V) Vopt 20
implanted in laboratory.
Both techniques are based on algorithms, which are Fig. 4. Experimentals points for G = 973 W/m2 andT = 24.6°C
implemented under MATLAB environment. Fig. 2
shows the block diagram of system under test. As can be observed there is a single MPP on the curve
Figure 3 presents the realized test bench corresponding to the optimum operating point for each
corresponding to this block diagram of the system. irradiation and temperature value.
In order to obtain the Ipv = f(Vpv) and Ppv = f(Vpv) This MPP corresponds to the maximum power Pmax,
curves, we proceed to a fast variation from 1% to 99% of which is associated to an optimum voltage, Vopt and
boost converter duty ratio (), by fixed step. Typical current, Iopt.
curves thus recorded, are in the form of those shown in To track this point, P&O and InC algorithms are
Fig. 4. implemented in MATLAB environment.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
117
S. Farhat, R. Alaoui, A. Kahaji, L. Bouhouch, A. Ihlal
L Io
The literature cites many mathematical models [14]
Vo
representing the highly nonlinear behavior of PVP,
Ce
caused by the photovoltaic effect of semiconductor PN MOSFET Cs Ro
Vpv Pulse
junction [15].
Of this literature review, we retain the complex
physical PV model which is represented by the
equivalent electrical circuit at one diode as shown in Fig.
5. The mathematical expression (1) reflecting the Fig. 6. Boost converter
electrical behavior of this model is given by:
The MPP is reached when Vpv = Vopt and Ipv = Iopt.
I pv I ph I d I sh Therefore, the optimum duty ratio is given by the Eq.
(5):
V pv Rs I pv Vopt
h G I o exp q 1 (1) opt 1 (5)
A ns K b T j
Ro I opt
V pv Rs I pv
Rsh Given 0 ≤ ≤ 1, the matching condition (6) is realized
when:
Different formula variables (1) are summarized in the Vopt
Ro (6)
nomenclature. I opt
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
118
S. Farhat, R. Alaoui, A. Kahaji, L. Bouhouch, A. Ihlal
20 the MPP [21]. Indeed, observing the Eqs. (7) and (8), the
Increase dP/dV>0
Vpv
MPP is reached when the slope of P-V curve is zero [22]:
10 Decrease
dPpv dI pv
Vpv Ppv V pv I pv I pv V pv (7)
dV pv dV pv
0
2 4 6 8 10 12 Vopt 1718 20
Vpv(V) dPpv dI pv I pv
0 (8)
dV pv dV pv V pv
Fig. 7. P&OMPPT
ΔP(k)=P(k)-P(k-
I pv I pv
for V pv Vopt At MPP
1) V pv V pv
I pv I pv
for V pv Vopt Left MPP (9)
ΔV(k)=Vpv(k)-Vpv(k- V pv V pv
1)
I pv I pv
for V pv Vopt Right MPP
V pv V pv
No Yes
ΔP(k) >
0 The implemented algorithm under MATLAB
environment, calculates dPpv/dVpv. This calculation is
based on the measurement of the PV voltage, and its
ΔV(k) > ΔV(k) >0 incremental output power.
0
If dPpv/dVpv is not close zero, the controller adjusts the
No No
boost converter's duty ratio, step by step until the PVP
Yes Yes
reaches its maximum output power.
α =α - Δα α =α + Δα α =α + Δα α =α - Δα
This implemented InC algorithm, is described by the
above flowchart in Fig. 9.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
119
S. Farhat, R. Alaoui, A. Kahaji, L. Bouhouch, A. Ihlal
Power (W)
20
30
ΔV=Vpv(k)-Vpv(k-
1) 28
14 15 16
0
0 5 10 15 20
Vpv (V)
ΔI=Ipv(k)-Ipv(k-1)
40
b
power (W)
31
20 30.5
No Yes 30
ΔV= 0 29.5
16 18 20
0
No Yes 0 2 4 6 8 10 12 14 16 18 20 21.5
Yes No
ΔI/ΔV= -I/V ΔI= 0 Time (s)
ΔI/ΔV> - ΔI > 0 40
I/V
30
power (W)
a
31
Yes No No Yes 20
30
α =α - Δα α =α + Δα α =α + Δα α =α - Δα 10
14 14.5 15 15.5
0
0 5 10 15 20
Vpv (V)
40
k=k+1
30
Power(W)
30.8
30.6
20
b 30.4
30.2
10 30
Fig. 9. Flowchart of the Incremental conductance algorithm 29.8
18 19 20 21
0
The used PVP for measurement is exposed to various 0 2 4 6 8 10 12 14 16 18 20 21.5
irradiance levels, and temperature, which vary with the Time (s)
climatic conditions, throughout the experimental
Figs. 11. Power (a) and response (b) by InC algorithm
measurements. The system control algorithm is for G = 973 W/m2 and T = 24.6 °C
implemented under MATLAB environment that
generates code for tracking the MPP in real time. Regarding the dynamic response variation of the
The results of the experimental measurements of the system, Figures 10(b) and 11(b) respectively show a
proposed InC technique will be presented and compared typical response of controllers P&O and InC, for given
with those to traditional P&O method. The experimental values of solar irradiation and temperature. Each of two
tests and measurements are carried by scanning the PVP studied system controllers evaluates the PVP output
electrical characteristic for different voltage levels power, and adjusts the power interface to satisfy the
ranging from 3 V to 19 V. This prevents the PVP to optimal operating conditions.
operate in short-circuit and open circuit. Further, the As shown in Fig. 10(b), in the case of P&O controller,
MPP is located inside this voltage interval. the MPP is achieved at steady state, after a little more
By performing a fast measurement, during than 2 s. While, as illustrated on Fig. 11(b), the MPP is
approximately 20 s, of Ppv = f (Vpv), under irradiance and reached in less than 2 s for the InC controller.
fairly uniform temperature conditions, we obtain the PVP During our investigations, the tests are carried for
powers shown in Figs. 10(a) and 11(a).Under the same solar irradiations varying from 600 W/m2 to 1000 W/m2
experimental conditions, we observe that for both approximately, and temperature values ranging from
proposed methods; P&O and InC, the obtained results 24°C to about 36°C. In these conditions, to evaluate the
converge towards the maximum power point and two studied techniques efficiency; P&O and InC, the
oscillate around this MPP. The PVP powers marked by instantaneous efficiency Effis calculated following the
dotted lines in these figures are obtained by varying duty formula (10). This last is the ratio between PMPPT point
cycle value of signal controlling the boost converter and the MPP.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
120
S. Farhat, R. Alaoui, A. Kahaji, L. Bouhouch, A. Ihlal
The PMPPT point represents the average value This situation is reversed when the MPPT operates
measured, at steady state, of the extracted power by the with a smaller step. So, a compromise between the
used techniques. The fact of considering this average dynamics and oscillations must be satisfied for the fixed
value allows eliminating the effect of measurement low step MPPT.
noise. For low irradiances levels we observe that the P&O
P efficiency is less than the InC technique. Contrariwise,
E ff 100 MPPT (10) for high irradiances levels, the P&O is more efficient
MPP
compared with the InC technique. We conclude that both
According to Fig. 12 which represents the obtained techniques; P&O and InC, are most efficient methods to
results concerning the efficiency Eff of the two techniques track the MPP. However, we can improve efficiency by
studied; P&O and InC, we estimate that these techniques developing a hybrid algorithm using these two methods
ensure broadly similar controls. and optionally with a variable step.
However, according to this Fig. 12, for low powers, The performances summary of studied MPPTs would
the MPPT of InC type is more efficient compared with serve as a useful guide to choose the appropriate MPPT
P&O. The situation is reversed for the power values method for engineers of photovoltaic industry.
greater than or equal to 28 W. However, as the evaluated
efficiency by (10), is in the range from 97 % to 99 %, we References
can conclude that both methods; P&O and INC, are
highly efficient for MPP tracking. [1] Kavitha, R., Thottungal, R., Cascaded multilevel inverter for
stand alone PV system with maximum power point tracking
technique, (2012) International Review of Electrical Engineering
100 (IREE), 7 (6), pp. 5939-5943.
[2] Khiari, B., Sellami, A., Andoulsi, R., Mami, A., A non linear
Efficiency (%)
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
121
Farhat, R. Alaoui, A. Kahaji, L. Bouhouch,
S. Farhat, Bouhouch, A. Ihlal
Proceedings of the 14th International Middle East Power Systems Abdelilah Kahaji Teacher of Electrical
Conference, MEPCON'10,
MEPCON'10, Cairo University, Egypt Egypt, December Engineering, with a Bachelor (1993) in
19-21,
19 21, 2010, pp. 750-755.
750 755. electronics of Higher Normal School of
[15] Weidong Xiao, W.G. Dunford, P.R. Palmer, A. Capel, Technology (ENSET), Mohammadia, Morocco.
Application of Centered Differentiation and Steepest Descent to Master of Business Administration (MBA)
Maximum Power Point TrackingTracking,, Industrial Electronics, IEEE (2002) at the University of Muncton Canada
Transactions on on, Vol. 54,Issue
54,Issue 5, 2007, pp. 2539
2539-2549.
2549. and a post graduate degree in (2005), of
[16] Henry Shu Shu-Hung
Hung Chung, K.K. Tse, S.Y. Ron Hui, Industrial Engineering in the National School of
C.M.Mok,M.T.Ho,A
C. M.Mok,M.T.Ho,A Novel Maximum Power Point Tracking Applied Sciences (ENSA) of Agadir, Moro
Morocco.
cco. His research, in the
Technique for Solar Panels Using a SEPIC or Cuk Converter,
Technique Converte context of national doctoral thesis, focuses on the thematic of
Power Electronics,
Electronics IEEE Transactions on, Vol.18,
Vol. Issue 3, 2003, Renewable Energies. The doctoral investigations took place in the
717 724.
pp. 717-724. Research Team in Advanced Technologies and Engineering of
[17] T. Esram, P.L. Chapman, Comparison of Photovoltaic Array Renewable Energies (ERTAIER) Agadir, Moroc
Morocco.
co.
Maximum Power Point Tracking Techniques, Energy Conversion,
IEEE Transac
Transactions
tions on,
on, Vol. 22, Issue 2, 2007, pp. 439-449.
439 449. Lahoussine Bouhouch Professorof
Professorof higher
[18] N. Femia, G. Petrone, G. Spagnuolo, M. Vitelli, Optimization of education at the ESTA (High School of
Perturb
erturb and O Observe
bserve maximum power point tracking method, Technologies of Agadir), IbnZohr University,
Power Electronics, IEEE Transactions on, on Vol. 20, Iss
Issue
ue 4, 2005, Agadir, Morocco. PhD Electrical Engineering at
963 973.
pp. 963-973. the Nancy I University, France in 1988 and state
[19] M.A.Eltawil, Z Z.Zhao,MPPT
Zhao,MPPT techniques for photovoltaic doctorate in Electrical Engineering in 2007.
applications
applications, Renewable and Sustainable Energy Reviews, Reviews, Vo ol. Responsible of the res
research
earch team ERTAIER
25, 2013
2013,, pp. 793
793-813
813. (Research Team inn Advanced Technologies and
[20] Chihming, Comparative study of peak
Hua Chihchiang, Shen Chihming, Engineering of Renewable Energies). His research focuses on topics
powertracking techniques for solar storage systems, Applied related to renewable energy, instrumentation and electromagnetic
Power Electr
Electronics
onics Conference
Conference and Expo Exposition,
sition, APEC
APEC'98.
'98. compatibility (EMC).
Conference Proceedings IEEE IEEE,, Thirteenth Annual
Annual, Vol.
Vol
2,Anaheim
Anaheim
Anaheim, CA,CA 15-19
15 19 Feb 1998, 697 685.
1998 pp. 697-685 Ahmed Ihlal, was born and brought up iinn
[21] D.P.Hohm
D.P. Hohm, M.E.Rop
Hohm, M.E.Ropp, p, Comparative Study of Maximum Morocco. He studied Physics and Chemistry
PowerPoint Tracking Algorithms Using an Experimental, and holds, in 1984, his BSc degree (Licence
(LicenceEs--
Programmable, Maximum Power Point Tracking Test Bed Bed, Sciences
ciences Physique)
Physique) in Solid
olid State
tate Physics
Physics from
Photovoltaic Specialists Conference, Conference Record of the the University Mohamed V, Rabat - Morocco
Morocco..
Twenty Eighth IEEE,
Twenty-Eighth IEEE, Anchorage, AK, 15--22 22 Sept. 20002000, He then joined Paris VII University–
University– France,
France,
1699 1702.
pp. 1699-1702. where he got, in 1985, a MSc. ddegreeegree (DEA:
[22] Issam Houssamo, Fabrice Locment, Manuela Sechilariu, Sechilariu Diplome des Etudes Approfondies) in Solar
Maximum power tracking for photovoltaic power system: Energy. He pursued his research on the studies and got, in 1988, a PhD
Development
evelopment and experimental comparison of two algorithms, degree from the University of Caen BasseNormandie - France. Dr. A.
Renewable Energy,
Energy, Vol.
Vol 35, Issue 10, 2010, pp. 2381-2387.
2381 2387. Ihlal started his teaching career on I988 as Assistant Professor in the
[23] Li Jiyong, Wang Honghua
Honghua, A n novel
ovel stand-alone
stand alone PV generation faculty of Science at University IbnZohr. Then he holds a
system based on variablestep size INC MPPT and SVPWM "Doctoratd'Etat" thesis in 1995. He is currently Full Professor in
control Powe
control, Powerr Electronics and M otion Control Conference,
Motion Faculty of Sciences, University IbnZohr, Agadir - Morocco. He is head
IPEMC'09, IEEE 6th International
IPEMC'09, International,, Wuhan, 17 17--20
20 May of the group working on ddeveloping
eveloping cost effective processes for the
2009, pp. 2155
2009,pp. 2155--2160
2160. fabrication of CIGS and CZTS absorber layers, buffer layers and TCOs.
He is working on PV and CSP systems as well. He has published 60
scientific papers, and acted as a referee for numerous international
Authors' information
Authors' journals. He has contributed to ththee organization of numerous national
and international conferences and was a member of scientific
Sadik Farhat Associate Professor ((2001) 2001) committees for several international conferences. He is supervising
inElectrical
inElectrical Engineering with a Bachelor (1996) PhD, MSc as well as BSc students in the field of PV and CSP. He is an
in electronics in Higher Normal School of expert of the CNRST iinn the field of renewable energies.
Technology (ENSET), Rabat, Morocco and a
post graduate degree in (2007), of Energy and
Environment in the National School of Applied
Sciences (ENSA) of Agadir, Morocco. His
research, in the context of national doctoral
thesis, focuses on the thematic of Renewable Energies. The doctoral
investigations took place in the Research Team in Advanced
Technologies and Engineering of Renewable Energies (ERTAIER)
Agadir, Morocco
Agadir, Morocco.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10,, N. 1
122
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – The application interval type-2 fuzzy inference systems (IT2FIS) has been attention
for a short-term load forecasting problems solution. This paper present optimization membership
function of antecedent (X,Y) and consequent (Z) interval type-2 Fuzzy Logic System using Big
Bang – Big Crunch Algorithm for application short term load forecasting on national holiday.
This method has implemented on the historical peak load data during 14 national holiday case
study in South and Central Kalimantan – Indonesia electrical power system on year 2008. The Big
Bang–Big Crunch (BB–BC) algorithm will applied to optimization interval footprint of uncertainty
(FOU) membership functions of interval type-2 fuzzy logic system. BB-BC Algorithm is a global
optimization algorithm and has a low computational cost and a high convergence speed, so very
efficient in this research. The test result showed main absolute percentage error (MAPE) less than
interval type-2 fuzzy inference system (FIS) and type-1 fuzzy inference system (FIS). Finally, this
paper defined main absolute percentage error (MAPE) 0.58045% for type-1 FIS, 0.53906% for
interval type-2 FIS, and 0.52421% for optimization interval type-2 FIS-Big Bang Big Crunch
Algorithm. Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved.
Keywords: Interval Type-2 Fuzzy Inference System, Big Bang - Big Crunch Algorithm, STLF,
MAPE
Nomenclature I. Introduction
BB-BC Big Bang – Big Crunch Load forecasting is very important in power system,
GA Genetic Algorithm because it could be estimate the power consumption at
IT2FIS Interval Type-2 Fuzzy Inference any given time. Accuracy in forecasting power loads can
System increase the security of power systems especially in the
IT2FLS Interval Type-2 Fuzzy Logic System operation of power systems such as power delivery (load
MAPE Main Absolut Percentage Error flow), maintenance unit and unit commitment.
MF Membership Function Therefore, the load forecasting can help in the
UMF Upper Membership Function reduction of operating costs and reliable case of electric
LMF Lower Membership Function power systems. Load characteristics of each area were a
STLF Short Term Load Forecasting different depending on the season and behavior of
à Type-2 Fuzzy Set society itself. Peak load on a special day such as a
Jx Primary membership of primary national holiday also has different characteristics from a
variable, x normal day for the particular holiday is always changing
FOU (Ã) Footprint of Uncertainty of IT2FS Ã either the day, date, month and season each year.
A x , A x Lower MF, Upper MF of IT2FS Ã Generally, peak load on national holidays tends to
decrease compared to normal days throughout the year.
yl Switch point left In this study, there are 14 national holidays were been
yr Switch point right
c used in short-term load forecasting on the electrical
xi Center of Mass power system of South and Central Kalimantan-
f Fitness function Indonesia.
l Upper and lower limit parameter Professor Zadeh from California University first
r Normal random introduced fuzzy logic in 1965.
k Step iteration Fuzzy logic describing mathematical calculations
xnew Upper and lower bounded based on set theory to describe the uncertainty in the
N Population Size form of linguistic variables, in another sense the theory
123
Akhmad Ramadhani, Agus Dharma, Imam Robandi
of fuzzy logic to develop set theory Boolean (0 and 1) be II.1. Interval Type-2 Fuzzy Set
the set of which has a membership value which is
An interval type-2 fuzzy sets (IT2FS) Ã symbolized
continuous and ambiguity (between 0 and 1) [1].
by membership function μà where x∈X and u∈Jx⊆ [0,
Professor Zadeh first introduced the concept of
1] characteristics could be recognize in the following
uncertainty of the type-2 fuzzy sets in 1975 as a
equation [4]:
development of the concept of ordinary fuzzy sets of
"fuzzy-fuzzy" or fuzzy-fuzzy type 1 are at right again [1].
Due to a system of type-1 fuzzy logic knowledge base A x,u
A xX xJ x,u Jx 0.1 (1)
is often used to build rules in a fuzzy logic system (FLS) x
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
124
Akhmad Ramadhani, Agus Dharma, Imam Robandi
k n N
Fuzzifier Defuzzifier
n 1
f y n n k 1 f n y n
Output yl mink1,N 1 n
Input k N
crisp x T1FS Y=f(x) n1 f nk 1 f n (7)
L n N
IT2 FS Inference Type- n 1
f y n n L 1 f n y n
Engine Reducer L n N
n1 f n L 1 f n
k n N n n
Rule Base n 1
f n y n k 1 f y
yr maxk1,N 1 k N n
n1 f n nk 1 f (8)
Fig. 2. Structure of Interval Type-2 Fuzzy Logic System (IT2FLS) R n N n n
n1 f n y n R 1 f y
R N n
II.3. Membership Function and Fuzzy Rules n1 f n n R 1 f
Fuzzy rules IF-THEN used in this method for
forecasting maximum load. where switch point of L and R:
In this paper input membership function (antecedent)
that is X ,Y and Output membership function y L yl y L 1
(consequent) is Z for short term load forecasting by (9)
R R 1
following by Eq. (5) as follows in: y yr y
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
125
Akhmad Ramadhani, Agus Dharma, Imam Robandi
x new xc lr / k (12)
YES
where xc stands for center of mass, l is the upper limit of
Define fuzzy rules interval type-2 fuzzy inference system
the parameter, r is a normal random number and k is the
iteration step. Then new point xnew was upper and lower Apply Operation of AND into interval type-2 fuzzy inference system
Apply Implication method MIN into FIS
bounded. The BB–BC optimization is approach takes the
Apply aggregation method MAX into FIS
following steps [8]:
Calculate defuzzification via type reducer using Kernik Mendel
Step 1 Form an initial generation of N candidates in a
Algorithm for define value of Forecast VLDmax
random manner. Respect the limits of the search
space.
Step 2 Calculate the fitness function values of all the Calculate Forecast Load Differences
Step 3 Find the center of mass according to Eq. (11). Calculate Error Forecast and Main Absolute Percent Error
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
126
Akhmad Ramadhani, Agus Dharma, Imam Robandi
Fig. 4. Single Line Diagram South and Central Kalimantan Electrical Power System
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
127
Akhmad Ramadhani, Agus Dharma, Imam Robandi
Peak Load Forecasted in year ith (MW) has calculated Indonesia electrical power system in year 2008 using
by Eq. (17) i.e.: optimization interval type-2 Fuzzy Inference System –
Big Bang Big Crunch algorithm is found on MAPE were
P'MAX MaxWD i relative smaller than interval type-2 FIS, and type-1 FIS
(17) method for load forecasting.
ForecastLDMAX ( i ) MaxWD( i ) This method forecasting using optimization interval
100 type-2 Fuzzy Inference System – Big Bang Big Crunch
algorithm have MAPE is less than 1.5%, This result was
The percentage of error is the difference between relatively small compare to standard MAPE permitted.
values with the calculation method that has real value.
The percentage of error had calculated by Eq. (18) i.e.:
P'MAX i MaxSD i
Error% 100 (18)
MaxSD i
V. Conclusion
Fig. 7. Membership function output variables Z
Application sort term load forecasting on national with optimization BBBC Algorithm
holiday case study in South and Central Kalimantan-
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
128
Akhmad Ramadhani, Agus Dharma, Imam Robandi
TABLE I
CALCULATION VLDMAX FROM YEAR 2005 UNTIL 2008
Name Holiday Year in 2005 Year in 2006 Year in 2007 Year in 2008
Type of Holiday
MaxWD LDMAX MaxWD LDMAX MaxWD LDMAX TLDMAX VLDMAX MaxWD LDMAX TLDMAX VLDMAX
1.New Year 211 -0.947 221.355 -6.5483 213.663 3.4716 -3.7481 7.21968 241.275 -0.52844 -1.3415 0.81308
Public Holiday
2.Independence Day 217.28 -5.2098 229.8475 -0.3078 242.838 1.2817 -2.7588 4.04056 236.5875 1.01971 -1.412 2.43169
1.Idul Adha 213.0225 -13.962 224.375 -11.8529 211.10 10.4453 -12.9075 23.3528 219.725 -3.2199 -5.1233 1.9033
Islamic Holiday
2.Islamic New Year 212.62 -5.0606 229.5975 -5.4998 222.77 -1.0504 -5.2803 4.2298 245.85 0.46776 -3.8703 4.33807
3.Maulid 224.945 -2.0382 221.3575 0.1231 242.793 -3.12303 -0.9576 -2.1655 249.1 -1.3649 -1.6794 0.31449
4.Isra Mi'raj 227.565 2.2345 235.51 -1.6602 243.563 1.98614 0.28715 1.69899 229.6625 4.23992 0.85348 3.38644
5.Idul Fitri I 208.515 -3.5129 229.145 -2.0751 234.675 -3.5262 -2.79402 -0.7321 235.625 -6.77984 -3.0381 -3.74177
6.Idul Fitri II 206.2725 -0.4084 226.4025 -3.4419 231.788 -1.8713 -1.9252 0.0538 231.5125 -5.815 -1.9072 -3.9078
1.Good Friday 216.475 1.7669 224.0625 0.41841 236.35 -1.9589 1.09268 -3.0516 247.25 1.19312 0.07547 1.11766
Christian Holiday
2.Ascention 224.3075 -5.0054 223.09 3.9222 241.475 0.01035 -0.5416 0.55196 241.9125 -4.8416 -0.3576 -4.484
3.Crismas 195.9425 4.2780 214.8825 3.14009 243.713 -1.9131 3.70906 -5.6222 227.75 5.59824 1.835 3.76324
1.Silent Day 223.1175 -1.7065 223.1175 -1.7065 231.068 -5.07535 -1.7065 -3.3689 218.4625 1.7566 -2.82945 4.58605
Hindus /Budhis
2. Imlek 216.5612 -8.3585 228.73 2.0417 185.398 5.0176 -3.1584 8.17599 240.7125 3.98297 -0.43306 4.41603
/Tionghoa
3.Waisak 222.5 1.9326 235.385 -1.5315 237.713 3.48635 0.20053 3.28583 242.5375 -2.03576 1.2958 -3.3316
TABLE II
FUZZY RULES BASE FOR LOAD FORECASTING IN YEAR 2008
Antecedent Consequent
Number of Rules
X Y Z
1 PVB PS PVS
2 PB PVS PS
3 PVB NB PS
4 PB PS PB
5 NS PB ZE
6 PS ZE PM
7 NVS PM NB
8 ZE NB NB
9 NM NB PVS
10 PVS PVS NB
11 NVB NB PB
12 NM PB PVB
13 PVB PVB PB
14 PM PB NM
TABLE III
FUZZY RULES BASE FOR LOAD FORECASTING IN YEAR 2008
Holiday Forecast Membership function degree of Fuzzy Set (μ) Rules
Type of Holiday Variabel VLDMAX
in year 2008 NVB NB NM NS NVS ZE PVS PS PM PB PVB (MAX)
Input X 1.69899 0 0 0 0 0 0 0.3010 0.6989 0 0 0 PS
Islamic
Isra Mi’raj Input Y 0.31448 0 0 0 0 0 0.6855 0.3144 0 0 0 0 ZE
Holiday
Output Z 3.38643 0 0 0 0 0 0 0 0 0.6135 0.3864 0 PM
TABLE IV
THE RESULT APPLICATION STLF USING IT2FIS-BBBC ALGORITHM WITH PREVIOUS METHODE IN YEAR 2008
Interval Type-2 FLS-BBBC
Type-1 FLS Interval Type-2 FLS
Name of holiday in Algorithm
Type of holiday
year 2008 P’MAX Actual Error P’MAX Actual Error P’MAX Actual Error
(MW) (MW) (%) (MW) (MW) (%) (MW) (MW) (%)
1.New Year 240.457 240 0.19042 240.4560 240.00 0.19002 240.447 240 0.1864
Public Holiday
2.Independence Day 237.974 239 0.42912 238.1220 239.00 0.36735 238.090 239 0.38051
1.Idul Adha 212.858 212.65 0.08193 212.8597 212.65 0.09865 212.8586 212.65 0.09803
Islamic Holiday
2.Islamic New Year 243.851 247 1.27481 244.4134 247.00 1.04717 244.2581 247 1.05424
3.Maulid 248.076 245.70 0.96742 247.9633 245.70 0.92118 248.1045 245.70 0.91246
4.Isra Mi'raj 232.696 239.40 2.80031 233.2612 239.40 2.56422 233.7479 239.40 2.34654
5.Idul Fitri I 219.038 219.65 0.27851 219.0401 219.65 0.27765 219.4659 219.65 0.27786
6.Idul Fitri II 217.833 218.05 0.09948 217.8344 218.05 0.09884 217.8393 218.05 0.09895
1.Good Friday 249.913 250.20 0.11439 249.9128 250.20 0.11479 249.908 250.20 0.11647
Christian Holiday
2.Ascention 231.364 230.20 0.50579 231.3650 230.20 0.50611 231.365 230.20 0.50642
3.Crismas 241.043 240.50 0.22601 241.0426 240.50 0.22563 241.043 240.50 0.22601
1.Silent Day 223.2004 222.30 0.40504 223.1840 222.30 0.39767 223.184 222.30 0.39786
Hindus /Buddhist
2. Imlek 249.302 250.30 0.39846 249.3012 250.30 0.39904 249.302 250.30 0.39885
/Chines
3. Waisak 238.404 237.60 0.33846 238.4041 237.60 0.33846 238.404 237.60 0.33846
MAPE 0.58045 0.53906 0.52421
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
129
Akhmad Ramadhani, Agus Dharma, Imam Robandi
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
130
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – The work presented in this paper concerns the conception, implementation and
testing of a photovoltaic system equipped with a new concept of controller 'regulator/ MPPT'. The
role of the controller is, firstly, to exploit the maximum power by Hill climbing improved algorithm
that considers optimal electrical characteristics of PV panels (variation of the optimal voltage of
the panels during operation around the maximum power point) regardless of the system
perturbation such as sudden variation of illumination. Secondly, to manage the energy between
panels and batteries and also, between panels and grid, in order to control the charge / discharge
process with taking into consideration the different situations that can occur in an installation
(charge cycle, discharge cycle, charge/discharge cycle). Thirdly, to inject the surplus of energy
provided by panels into the grid when the batteries are fully charged and even when they are in
charge phases (second and third phases of charge while regulation is enabled; a part of the power
delivered by the panels is not absorbed by the batteries).
And also, its role is to estimate the state of charge with precision by the combination of two
methods: the OCV method and the coulomb counting method. In addition, the system ensures all of
the following functions: display of parameters on LCD, acquisition and presentation of results on
computer, supervision and so on.
The results obtained during days of experimentation, under weather disturbances (variation of the
irradiance and temperature) show the robustness of the PV system, the good control of batteries
charge/discharge process under the maximum of energy provided by the PV panels and the
protection against overcharging and deep discharging. The surplus of energy provided by panels
and unabsorbed by the batteries is transferred into the grid. The estimation of the state of charge
SoC is evaluated from the measured parameters in real time; it shows an improvement of around
5% compared to the conventional technique. It is conceived so that it can be adapted to any
battery system of type lead. Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved.
131
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
In general, they are not precise and have their climbing algorithm, taking into consideration the optimal
limitations [8]-[10]. electrical characteristics (voltage, current) of cells and
These methods include: PV panels, this algorithm presents are markable rapidity
The specific gravity method [10] and the method at the level of convergence towards the MPP point
based on the measurement of the open circuit voltage compared to the classical algorithm [15]-[16].
(OCV) [8]. Both methods are not practical for an The determination of the SoC is achieved by the
online estimation since measurements are made after combination of two methods to correct estimation errors
a relatively long period of stabilization and and to increase precision, taking into consideration the
relaxation. faradic efficiency of charging/discharging [8]. The
Coulomb counting method based on the integral over initialization of SoC is fixed by the method of open
time of the current entering (charging) or leaving circuit voltage (OCV) [8] when the batteries are at rest: a
(discharging) the cells [9]. The precision of this first recalibration of the state of charge is applied.
method is closely related to the techniques of The quantity of charge absorbed or delivered by the
measurement and integration of the current, which is batteries is calculated by the Coulomb counting method
variable and depends on weather conditions [8]. in real time [9] and a second recalibration of the state of
The method of measuring the internal impedance of charge is applied when the batteries reaches full charge
the battery is not reliable since the impedance is conditions. The proposed method is characterized by its
temperature, acid density and ageing dependent, simplicity, it is easy to implement and does not require
which increases the complexity of using it [11]. the use of additional hardware. So, the SoC is estimated
Photovoltaic systems connected to the grid [12] have with precision in real time. The proposed power system
for the main role the injection of energy supplied to the is designed in order to manage the energy flow and
grid. Among their structures, one that integrates elements optimize the system functioning between panels and
of storage (batteries) [12]-[13] either to satisfy needs of battery as well as between panels and grid. This power
local energy demand [14] or to facilitate coupling block is controlled by PWM signals generated by a
(presence of intermediate) and to resolve problems of microcontroller. The various measurements and
adaptation between production and demand [12]-[14]. parameters will be displayed on an LCD and transferred
The main challenges for these structures: through a serial connection to a computer [6].
Protection of batteries used in the process charge /
discharge
The stability and energy storage when the batteries II. Structure and Functioning
are fully charged of the System
The operating performance of the energy supplied
II.1. Synoptic Diagram
during the charging phases (part of delivered energy
is not absorbed during the phases of battery charging. The synoptic diagram of the PV system designed and
This paper proposes and studies a structure of a tested in this work (Fig. 1) consists of various blocks and
photovoltaic system equipped with a digital MPPT functional subsystems:
regulator which guarantees the storage of the energy Two PV panels mounted in parallel [6].
produced by the PV panels in lead acid batteries, under Two lead acid solar batteries in series.
optimal conditions, with a good precision of the state of DC or AC load through a DC/AC converter (lamp,
charge estimation. This structure maximizes the power pump, ...)
transferred to the converter output, increases the rapidity A DC/DC Boost (or Buck) converter to adapt the PV
of charging and improves efficiencies compared to some panels to the batteries (functioning around the
commercial regulators. A new concept is proposed to maximum power point) [17].
control and manage the energy supplied by panels in the The management and supervision system (MSS),
charging/discharging process by a regulation system. which is based on the use of a microcontroller to
This latter is based firstly on the hash of current perform all the tasks of: optimization of the system
supplied by panels via a PWM signal controlling the functioning by the MPPT control, regulation of
switch Buck or Boost DC/DC converter to maximize the charge/discharge process, control of the energy
power to the output by the execution of an algorithm that transfer between various blocs, acquisition and
performs the function MPPT. And secondly, it is based display of different quantities (voltages, currents,
on the hash of current supplied to the output of the powers, SoC, performance, illumination, temperature)
converter to charge the batteries with a second PWM on the screen of a PC or LCD and communication to
signal controlling the charge circuit. A part of energy the computer by a serial link [18].
produced and unabsorbed by the batteries during Block of energy flow controller, formed by three
regulation (batteries charge) is injected into the grid with power circuits (circuits 1 to 3) presented in Fig. 1, that
the minimum of energy loss by using a power circuit control the charge/discharge process of the batteries and
controlled via a third PWM signal. When the batteries are inject the surplus of energy into the grid with the
charged, all of energy produced is injected into the grid. minimum of energy loss either during the charging
The MPPT control is ensured by an improved Hill phases, or when the batteries are fully charged.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
132
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
133
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
134
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
When the voltage OCV is 25 V and the temperature controlling the DC/DC converter. The second one
increases (decreases) 15 ° C around T = 25 ° C (40% (PWM1) controls the switch 1 (circuit 1) in order to
variation) then SoCi increases (decreases) to 83.6% control and manage the batteries charge. And the third
(76.4%) (a variation of 3.6%). one (PWM2) controls switch 2 (circuit 2) to supervise
and manage the transfer of energy surplus produced into
the grid. The discharge of the batteries is provided by the
circuit 3controlled by a square signal (S.C.D). The
overalls results show:
During the functioning of the complete PV system,
The MSS system generates the PWM signal (Fig. 8)
of a duty cycle, for this sample, coincides with a
value of 0.7. This allows adapting the PV panels to
the load (Batteries and grid) and extracting the
maximum of power from PV panels by following the
point MPP.
Fig. 5. The proposed method to estimate the state of charge During the charging cycle:
SOC during the charge/ discharge of the batteries
During the Boost phase (Fig. 7(a)), the batteries
120 are charging with IBAT current. However the VBAT
voltage and the state of charge SoC increase
100
linearly with time. During this phase, the
80 management system MSS generates the PWM1
SoCi (%)
60
signal (Fig. 9) of a duty cycle of 0.98, to activate
correctly the power circuit 1 (circuit 1 in Fig. 1).
40 This latter controls the batteries charge and the
SoCi (0°C)
20 SoCi (25°C)
signal PWM2 (Fig. 9) of a 0.02 duty cycle
SoCi (50°C) (inverse of PWM1) to deactivate correctly the
0
power circuit 2 (circuit 2 in Fig. 1) which controls
23,0 23,5 24,0 24,5 25,0 25,5 the transfer of power to the grid. During this
OCV (V) phase, the power supplied by the PV panels is
Fig. 6. The initial state of charge SoCi based on the open circuit voltage
OCV for a 24V battery at temperatures of 0°C, 25°C and 50°C
completely absorbed by the batteries, until a
voltage VR that is limited to 28.4V in order to
protect batteries against overcharging. At this
III. Simulation Results voltage, a capacity of 85 to 90% is recovered and
the batteries are relatively charged that allows the
To simulate the various blocks of the system, the PV system to generates a S.C.D signal in state 1 (state
system of Fig. 1 is implemented as blocks in ON) to control circuit 3 in order to discharge the
ISISPROTEUS [24], and then simulated the system batteries if a load is connected.
operation taking into consideration the algorithms of Fig. During the absorption phase (Fig. 7(a)), the
2 to Fig. 5. In the corresponding program where all voltage VBAT is limited around VR=28.4V. The
instructions, equations defined above and functions that SoC continues to increase gradually depending on
as sure the various tasks are programmed and injected the power absorbed that decreases, due to the
into a micro controller which plays the role of a system decrease of the charging current, to achieve full
processor. The parameters related tour system: panel charge (VBAT = 28.4V and SoC> 95%). In this
type, batteries (Table I), acquisition,... are set to optimize case, the management system MSS generates two
the operation, and then extracted the graphics that PWM signals. The first signal (PWM1) (Fig. 10),
illustrate the different operating modes: functioning of of a variable duty cycle (0.02 to 0.98) is to adjust
the MPPT control, control of energy flow, and estimating the VBAT voltage around VR, by using pulses of
the state of charge. The results of all possible modes of current generated by the hashing of the current
the batteries operation, when the energy is transferred IBAT through the power circuit1. And the second
into the batteries (charging cycle, discharging cycle and signal (PWM2) (Fig. 10), of a variable duty cycle
charging/discharging cycle), are presented in Fig. 7(a) (0.02 to 0.98) which is practically the opposite of
and Fig. 7(b). They show the typical evolution of the PWM1,is to activate or deactivate correctly the
voltage VBAT, the state of charge SoC, the charge current power circuit 2 to inject the surplus of energy into
IBAT and the discharge current Idech during different the grid.
cycles. During this process, the management and During this phase, the charging of batteries
supervision system controls the different power circuits continues through the current pulses IBAT
(Fig. 1) by PWM signalsof10kHzfrequencyandvariable generated by the PWM1 until a SoC which
duty cycle to ensure the good functioning of the system. exceeds 95% and a current IBAT less than IOCT. The
The Fig. 8 to Fig. 12 extract three PWM signals. A first surplus (energy supplied) is transferred into the
one (PWM) ensures the MPPT function of the system by
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
135
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
grid via the circuit of power 2 controlled by the conditions, a power of 65W at a voltage of 14.2V and
PWM2 signal. At this stage, the batteries are fully a current of 4.4 A. [6]-[25].
charged and the system continues to generate the Two lead acid solar batteries (flat plat) mounted in
S.C.D signal in state 1 (state ON) to control series. Each battery has a nominal voltage and
circuit 3 in order to discharge the batteries if a capacity of 12V and 110Ah.
load is connected. A DC / DC boost converter dimensioned to operate at
During the floating phase (Fig. 7(a)), the charging a frequency of 10 kHz, a power of about 200 W and a
current decreases above IOCT and the SoC is over current ranging between 1A and 10A [6].
95 %. The batteries keep on charging through A management and supervision system (MSS) (Figs.
current pulses to stabilize the voltage VBAT around 14), which provides communication between the
VFLT=27.3V. During this phase, the management different blocks of the PV system. It processes the
system MSS generates the PWM1 signal (Fig. data, acquires and displays all the electrical
11), of a variable duty cycle around 0.02 to parameters and controls the switches of the electronic
activate correctly power circuit 1, which controls circuits. This ensures the good functioning of the
the charge. A low energy supplied by the PV MPPT control, the regulation and control the energy
panels is transferred to the batteries in order to flow, the estimation of the state of charge and the
compensate the self-discharge, and almost all of management and drives a LCD display. A serial link
the energy is injected into the grid. Since the is used to transfer, present and storage data into a
batteries are fully charged, the discharge may be computer. The application developed in this work
performed if a load is connected via power circuit applies to all stand-alone PV systems equipped by
3, which is controlled by the S.C.D signal lead acid batteries. The algorithms 2 to 4 may be
generated by the MSS system, to the highest state adapted for the system (choice of components, panels
(state ON). and batteries’ characteristics, conditions
During the discharge cycle (Fig. 7(a)), the square charge/discharge.) so that the whole installation is
S.C.D signal controls power circuit 3 to supply the reliable and optimized.
load from the batteries (Fig. 12). The voltage VBAT
and the SoC decrease gradually to the threshold
SoCmin.D set by the algorithm of discharge (Fig. 4).
In addition, the S.C.D signal changes to 0 (state OFF)
because the SoC reached SoCmin.D in order to stop
the discharge. To initiate battery charging, the system
generates a PWM1 signal with a duty cycle of the
order of 0.98 that controls power circuit 1 in order to
transfer all the energy into the batteries.
During the tests of the batteries operating modes:
charge cycle, discharge cycle and charge/discharge
cycle, when the charge is carried by IBAT = 2.3A and
the discharge by Idech = 5A, the results are shown in (a) the charging phases and simple discharge
Fig. 7(b):
During the charge cycle, VBAT and SoC increase
progressively until SoC = 70%.
During the discharge cycle, VBAT decreases and so
does the SoC until reaching the minimum state of
charge (SoCmin.D); then the discharge is
disabled.
During charging/discharging cycle, the MSS
system activates the discharge when a 5% of the
capacity is recovered (SoC = SoCmin.R = 45%).
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
136
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Fig. 9. Samples of the: S.C.D, PWM1and PWM2 signals IV.2. Results and Discussion
during the Boost charging phase
To validate all the simulation results, the PV system
designed and carried out in this work (Fig. 13) is tested
in the normal functioning of installation (production of
energy, batteries charge, injection into the grid) during 8
days (D1 to D8) successive and globally illuminated in
order to fully charge the batteries.
Fig. 10. Samples of the: S.C.D, PWM1 and PWM2 signals during the
Absorption charging phase
(a)
(b)
(c)
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
137
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Copyright © 201
2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10,, N. 1
138
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Fig. 15. The typical illumination variation during the days Figs. 19. Samples of the signals: (a): PWM1 of the duty cycle of 0.68
the experimentation took place controlling the batteries charge, (b): PWM2 of the duty cycle of 0.32
controlling the injection of energy into the grid
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
139
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
140
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Tcla / Timp
2,0
the power absorbed (injected) by the batteries (grid). 2
700 W/m
The global efficiency of the PV system during the 1,5
2
940 W/m
charging and the injection into the grid (Fig. 27) is 1,0
(c)
V. Conclusion
Figs. 24. The evolution of: (a): experimental and simulated (Optimum)
In this paper, the design, implementation and results of Vpv voltage; (b): experimental and simulated (Optimum)
results of Ipv current; (c): experimental and simulated (Optimum)
functioning of a photovoltaic system are studied. results of Ppv power
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
141
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Acknowledgements
This work is supported by:
Belgian Development Agency CTB (Project
MIP/012/010), Morocco.
United Nations Development Programme UNDP Art
Gold Morocco, ENV 2008 2 OO.
Cooperation Moroccan-Belgian "Institutional
University Commission", IUC, Oujda, 2008-2012
(Water and Environment Activity / Sub-Activity
Renewable Energy).
Moroccan-Tunisian Cooperation, SCIENTIFIC
Fig. 28. The evolution of the powers provided by the system conceived RESEARCH AND TECHNOLOGY Project
(Pcnv) and a commercial regulator (Preg) during the battery charging
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
142
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
143
K. Hirech, M. Melhaoui, E. Baghaz, R. Malek, K. Kassmi
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
144
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Abstract – For economic and ecological considerations and following reasons for the depletion
of certain oil resources, the coupling of renewable energy sources with the production of hydrogen
by electrolysis of water contributes to the development of intelligent networks called "Smart Grid"
(SG). In this context, this paper aims to develop an algorithm for the management and control of a
hybrid microgrid by multi-agent system (MAS). This hybrid Microgrid is composed of a
photovoltaic source (PVECS), a fuel cell (PEMFC), an electrolyzer (ELZ), and a hydrogen storage
system (H2TANK). Oversight of this microgrid is under the control of MAS and ensures optimum
performance, taking into account constraints such as intermittent renewable energy, supply /
demand balance and the price per kWh. Copyright © 2015 Praise Worthy Prize S.r.l. - All rights
reserved.
145
Lahoussine Elmahni, Lahoussine Bouhouch, Rachid Alaoui, Ali Moudden
of energy production from renewable sources and The contribution of our work is to model and develop
particularly those based on hydrogen, which will become an approach to the management of a microgrid using
in the near future an essential solution to energy and MAS.
environmental problems. In addition, the massive Our paper is organized as follows: after the
proliferation of these distributed resources and the arrival nomenclature and the introduction, section 3 outlines the
of electric vehicle require innovative and intelligent context and work tools. Section 4 describes each physical
solutions for the control and management of the and software component of the microgrid. In section 5
electricity grid. This intelligent control of power systems the focus will be on the control and management of the
is called "Smart Grid" (SG) [1]. network by MAS. In section 6, we present the results of
The balance between production and consumption is simulations, as well as the interest and the reliability of
provided by SG implementing parallel systems of the chosen configuration. Finally a conclusion and a list
communication and integrating distributed information of references will complete our paper.
networks.
Recently, control and management of MGs by MAS
[2] have indeed attracted the attention of many II. Context and Work Tools
researchers, given the perfect concordance of their II.1. Microgrid Subtest
characteristics with those of MGs. Indeed:
- The distributed nature gives the MAS system the During our investigations, we are interested in
ability to adapt well to SG. studying the microgrid (MG) ensuring the supply of a
- The autonomy of the MAS: MAS must take proactive village of 1300 people located 50 km north of Agadir.
steps in seeking to meet their needs based on their For this village, the daily consumption is estimated at
roles and their limits. For example in the case of a 330 kWh. An extract from the daily consumption is
charge that seeks to be powered via a source, MAS presented in Fig. 1; where you notice a peak
are trying to minimize costs while maintaining consumption of 24 kWh goshawks of 20 h [17].
operational constraints. Meteorological data concerning this village are extracted
- The opening of the system: MAS easily adapt to from PVGIS (Photovoltaic Geographical Information
changes in the network, even in the event of a fault, System).
an addition or removal of elements. This procedure is The model that we have developed in this work is
ensured even if these operations were not included in designed to optimize production and consumption using
the system design. the microgrid configuration shown in Fig. 2.
The application of multi-agent systems is also LOAD PROFILE
developed for solving optimization problems [3], the 25
restoration of the initial state of MG [4], the integration
of electric vehicles [5] and cyber security in smart Grid
Power (Kw)
20
[6].
According to some authors of scientific literature,
several areas implement management based on the MAS. 15
For example, Dimeas [7] introduced a general
framework for control of microgrid based on technology
of MAS. Pipattanasomporn [8] has proposed a 10
0 5 10 15 20
framework for the response to the request that can be Tim e (h)
implemented within a network of an intelligent House.
Logentthiran [9] introduced the operation in real time of Fig. 1. Typical load profile
a microgrid using MAS. Nagata [10] introduced a
method to keep the system voltage in the optimal range GRID
in order to avoid voltage instability phenomena using AC
multiple agents. Wang [11] proposed models of fuel
cells, wind turbines and photovoltaic-based systems.
Muhammad Ali [12] studied architecture which INVERTER
LOAD AC/DC
upgrades the existing system by monitoring all control 330 KWh/d
signals through which they pass between production DC
units, servers, connected loads and the protective devices
by using the TCP/IP. FUEL CELL
On modeling and simulation of decentralized 25 KW
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
146
Lahoussine Elmahni, Lahoussine Bouhouch, Rachid Alaoui, Ali Moudden
Fig. 3. Interfacing provided by MACSimJX III.1.3. Hydrogen Storage System Model (H2TANK)
One of the hydrogen storage techniques is physical
III. Microgrid Configuration hydrogen storage which is to store hydrogen in the form
of compressed gas or liquid.
III.1. Description of the Physical System The difference between the amount of hydrogen
The study articles of scientific and technical literature, produced by the ELZ and consumed by the PEMFC is
mentioned in the introduction, for modeling and sent to the storage tank.
simulation of decentralized renewable energy systems, In this study, the model of hydrogen storage is based
led us in the remainder of this document, to be limited to on Eq. (10) [23] which directly calculates the tank
simplified models to accelerate code execution time, pressure using the ratio between the flow of hydrogen
under the Matlab environment. The models we have and the volume of the tank:
chosen will be described below.
− = (10)
III.1.1. Model of PVECS
Fig. 4 summarizes the electric model chosen for the
photovoltaic cell whose Eqs. (1)-(5) characterized the
behavior [21]:
= + + (1)
= [ − 1] (2)
= + (3)
+
= − [ − 1] − (4) Fig. 4. Model of the photovoltaic cell
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
147
Lahoussine Elmahni, Lahoussine Bouhouch, Rachid Alaoui, Ali Moudden
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
148
Lahoussine Elmahni, Lahoussine Bouhouch, Rachid Alaoui, Ali Moudden
Indeed, these rated powers PL, PLC, PLNC, PG, PV, PELZ
and PEMFC are respectively the power load, critical load,
non-critical loads, network (GRID) of photovoltaic
(PVECS ),the electrolyzer and the PEMFC. These
powers are linked via relationship (11) and (12):
+ + = + (11)
= + (12)
V. Simulation Results
The MAS system developed with JADE at base of the
flowchart in Fig. 12 and the physical system under
SIMULINK (Fig. 11), has enabled us to develop a
control strategy using collaborative agents for the overall
task of the system.
All statements submitted subsequently relate to July
and November 2012.
TABLE I
BLOCKS MODELING MG UNDER STUDY
Fig. 10. Graphical user interface of the agent sniffer jade Bloc Model
1 Meteorological data
2 Photovoltaic panels
III.3. Global Model 3 Power inverter
4 Filter
The global model that we developed under Simulink is 5 Step-up transformer 240 V/25 kV
represented on Fig. 11. It is composed of a set of blocks 6 Grid
that Table I summarizes. 7 Loads
8 Step-down converter or series (Buck)
9 Electrolyzer
10 Hydrogen tank
IV. Management Algorithm Microgrid 11 Truck or parallel converter (Boost)
12 PEMFC
Managing the flow of energy in the SG, according to a 13 Interfaces measurement
control algorithm, provides smart features; eg load 14 S-FunctionMACSimJX
management and operation of island mode (autonomous). 15 Breakers
It is therefore interesting to bring equation, and ignoring 16 Visualization Tools
losses, power flows involved in the SG.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
149
Lahoussine Elmahni, Lahoussine Bouhouch, Rachid Alaoui, Ali Moudden
Start
Read powers
Pv, PLC, PLNC, PEMFC, PELZ, PG
YES NO
PV ≥ PL
NO YES
PL>PV ≥ PLC
100 % working PV
PEMFC= 0 ;PG< 0 PEMFC =0 ;PELZ = 0 ; PG= 0
PV<PLC
YES Load Management
NO
Load Management
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
150
Lahoussine Elmahni, Lahoussine Bouhouch, Rachid Alaoui, Ali Moudden
140
H o u r l y G l o b a l S o l a r R a d i a ti o n ( k w /m 2 )
1.4
JULY NOVEMBER 120
1.2 PVEC S
PEMFC
100
1 EL Z
Po w e r (KW)
80 L OAD
0.8
60
0.6
40
0.4 20
0.2 0
0 4 8 12 16 20 23
0 Ti m e (h )
0 4 8 12 16 20 23
Ti m e (h ) Fig. 17. Evolution of daily power for the July 1st
14
Fig. 13. Global horizontal radiation daily in July and November
November
12 july
PVECS PEMFC ELZ LOAD
120 10
Power (KW)
100 8
80 6
Po w e r (kw )
60 4
2
40
0
20 0 20 40 60 80 100
Ti m e (h )
0
0 23 47 71 95 119 Fig. 18. Daily power injected to the grid in July and November
Ti m e (h )
12
Fig. 14. Evolution daily power in November
Novem ber
P o w e r s o l d to g r i d ( K W )
80 4
60 2
40
0
0 5 10 15 20
20
Tim e (h)
0
0 23 47 71 95 119 Fig. 19. Daily power injected to the grid the first day
Tim e (h)
of November and July
20
120
100 PVECS
15
PEMFC
80 ELZ
P o w e r (kw )
LOAD 10
60
40
5
20
0
0 0 20 40 60 80 100
0 4 8 12 16 20 23
Tim e (h) Tim e (h)
Fig. 16. Evolution of daily power for the November 1st Fig. 20. Quantity of hydrogen in the tank in Novemberand July
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
151
Lahoussine Elmahni, Lahoussine Bouhouch, Rachid Alaoui, Ali Moudden
Fig. 21 matches the 1st day of the months July and System for Demand Side Management in Smart Grid, Power
Electronics and Drive Systems (PEDS), IEEE2011 Ninth
November. This curve shows that the hydrogen is stored
International Conference on, Singapore 5-8, Dec. 2011, pp. 424-
when there is an excess of solar energy, that is to say, 429.
between 7 AM and 5 PM. [4] T. Logenthiran, D. Srinivasan, A.M. Khambadkone, H.N. Aung,
This hydrogen is re-used in the opposite case, that is to Scalable Multi-Agent System (MAS) for Operation of a
Microgrid in Islanded Mode, Power Electronics, Drives and
say, between 5 PM and 7 AM. It is also important to note Energy Systems (PEDES) & 2010 Power India, IEEE2010 Joint
that the hydrogen tank is never empty, which can meet International Conference on, New Delhi 20-23, Dec. 2010, pp. 1-
the demand of consumers. 6.
[5] Shao Shengnan, M. Pipattanasomporn, S. Rahman, Grid
200 20 Integration of Electric Vehicles and Demand Response With
July Hydrogen Tank level Customer Choice, Smart Grid, IEEE Transactions on, Vol. 3,
Issue 1, 2012, pp. 543-550.
Novemeber Hydrogen Tank level
H yd ro g e n Ta n k L e ve l (kg )
15 [6] W. Y. Wang, Z. Lu, Cyber security in the Smart Grid: Survey and
July PV Power challenges, Computer Networks (Amsterdam, Netherlands :
PV Po w e r(kw )
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
152
Lahoussine Elmahni, Lahoussine Bouhouch, Rachid Alaoui, Ali Moudden
Cell - Electrolyzer Micro-Grid Modeling and Control in Real My Rachid El Moutawakil Alaoui Professor at
Time Digital Simulator, Clean Electrical Power, IEEE2009 the University Ibn Zohr Agadir,Morocco, with a
International Conference on, Capri 9-11 June, 2009, pp. 29-34. Bachelor in Electronics ENSET, Rabat and
[23] M. Uzunoglu, O.C. Onar, M.S. Alam, Modeling, Control and Masters in Industrial Engineering at the ENSA
Simulation of a PV/FC/UC based Hybrid Power Generation Agadir. His doctoral thesis focused on the
System for Stand-Alone Applications, Renewable Energy, Vol. Defensive Software processors. Permanent
34, Issue 3, 2009, pp. 509-520. member of the Research Team of Advanced
[24] Ciresan, A., Draghici, D., Gurbina, M., Lascu, D., A new boost Technologies and Engineering Renewable
DC-DC converter exhibiting low stresses and high efficiency, Energy ERTAIER. Responsible for electromagnetic compatibility
(2013) International Review of Electrical Engineering (IREE), 8 (CEM) research focus.
(6), pp. 1694-1700.
[25] Ajaamoum, M., Kourchi, M., Bouachrine, B., Ihlal, A., Pr. Ali Moudden was born in 1958 in Morocco.
Bouhouch, L., Photovoltaic panel emulators, design and He obtained (1987) the "Doctorat d'état" in
implementation using rapid prototyping technique, (2014) physics at the University of Caen (France). He
International Review of Electrical Engineering (IREE), 9 (5), pp. founded with other colleagues in 1991, the
1012-1020. laboratory of instrumentation and measurements
at the Faculty of Sciences of Agadir (Morocco),
that became in 2005 the laboratory of metrology
Authors’ information and information processing. Laboratory
activities are focused mainly on the development of physical techniques
Lahoussine Elmahni Was born in Agadir, for materials evaluation and processes monitoring. He was director of
Morocco in 1968; Professor (1993) in electrical higher school of technology of Agadir during the period 2005-2011.
engineering with bachelor(1989) in electronics
in higher normal school of technology
(ENSET), Rabat and a post graduate degree in
(2008), of energy and environment in the
National School of Applied Sciences (ENSA) of
Agadir, His research is focused on Smart Grid,
electric vehicles, demand response, energy efficiency, renewable
energy integration, energy storage and distributed resources.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
153
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Load Frequency Control of Multi Area SSSC and CES Based System
Under Deregulation Using Particle Swarm Optimization
Abstract – This paper deals with the application of various evolutionary optimization techniques
for designing the gain of integral controller in a multi area hydrothermal system under
deregulation.
The various techniques which have been used are Simulated Annealing (SA), Genetic Algorithm
(GA) and Particle Swarm Optimization (PSO). A two area hydrothermal system in the presence of
Static Synchronous Series Compensator (SSSC) and Capacitive Energy Storage (CES) under
deregulation is considered to exemplify the optimum parameter search. The integral of square of
error (performance) index is considered in the search of optimal Automatic Generation Control
(AGC) parameters. The results presented in this paper demonstrate the superior working of PSO
over others in the tuning of AGC. Copyright © 2015 Praise Worthy Prize S.r.l. - All rights
reserved.
154
P. Marimuthu, B. Basavaraja, Subhransu Sekhar Dash
II. System Investigation Nominal parameters of the system are given in the
Appendix.
The AGC system investigated is composed of an The performance index (PI) namely, integral of square
interconnection of two areas. Area 1 comprises of a of error (ISE) given by:
reheat system and Area 2 comprises of hydro system.
The detailed transfer function models of speed t
f1
2
governors and turbines are discussed and developed in J f 22 Ptie2 12 (1)
the IEEE committee report on Dynamic models for 0
Steam and Hydro turbines in Power systems [12].
The detailed small perturbation transfer function block is considered in this work to compare the performance of
diagram model of two area hydrothermal system under various optimization methods.
open market scenario is shown in Fig. 1.
Cpf11 Cpf12
+ +
Cpf21 + Cpf22
p.u load of + p.u load of
Disco 1 Disco 2
Cpf31 Cpf32
+ +
Cpf41 + + Cpf42
B 1 1
R1 R2 PD1(s )
Thermal area
PCES
ACE
+ - + -
1 1 1 sK r Tr + - K p1
Ki/s a
p - 1 s Tg 1 s Tt 1 s Tr 1 s T p1
+ +
+
+
- +
Demand of
1 1 1 sK r Tr discos in
a area 1 to
p - 1 s Tg 1 s Tt 1 s Tr Gencos in
+ area 2
+ +
+ 2 T12
+ s
- PSSSC -
+ + Demand of
-1 discos in
- 1 1 sTR 1 - sTw
a area 2 to
p 1 s T1 1 s T2 1 0.5s Tw Gencos in
-1 area 1
-
+ + + +
- 1 1 sTR 1 - sTw KP2
Ki/s a
1 s T1 1 s T2 1 0.5s Tw + - 1 s TP 2
p
+ -
-
ACE + Hydro area
PCES
1 1 PD 2 (s )
B R3 R4
Cpf13 Cpf14
+ +
Cpf23 + Cpf24
p.u load of + p.u load of
Disco 3 Disco 4
Cpf33 Cpf34
+ +
Cpf43 + Cpf44
Fig. 1. Two Area SSSC and CES based Hydrothermal System under Deregulation
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
155
P. Marimuthu, B. Basavaraja, Subhransu Sekhar Dash
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
156
P. Marimuthu, B. Basavaraja, Subhransu Sekhar Dash
From Eq. (7) the term cos m c can be written as: Pmn Ptie PSSSC
which implies:
Vn
cos m c cos n c (8)
Vm VmVn
X sin mn
T
Referring to Fig. 4 it can be written as: PSSSC (15)
Vs
yw V V 2V V cos
2 2
cos n c (9) m n m n mn
xy
Based on Eq. (15) it can be observed that the power
and it can be seen as: output of SSSC can be controlled by varying the SSSC
voltage Vs and thus the frequency and tie line
yw Vm sin mn (10) deviations can be controlled. Fig. 5 shows the structure
of SSSC to be incorporated along the hydrothermal
also: system. It can be seen from Fig. 5 that the structure of
xy Vm2 Vn2 2VmVn cos mn SSSC consists of gain block K SSSC , time constant TSSSC
(11)
and two stage phase compensation blocks having time
and mn m n
constants T1 , T2 , T3 , T4 respectively.
Using these relationships Eq. (11) can be modified as:
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
157
P. Marimuthu, B. Basavaraja, Subhransu Sekhar Dash
energy distributed probabilistically according to Eq. (16): that are highly adapted to their environments. GA is an
algorithm that operates on a similar principle.
E
P E exp (16) Algorithm:
kT
Initialization of binary chromosome strings of
population, each consisting of the gains.
where k is the Boltzmann constant. Therefore by
Decoding of strings and calculation of the fitness
controlling the temperature T and assuming that the
function.
search process follows the Boltzmann probability
Selection of strings into the mating pool based on
distribution, the convergence of the algorithm can be
Roulette wheel selection criteria
controlled.
According to Metropolis et al. (1953) if at any instant Performing the Crossover and mutation to generate
off springs.
the current point is xt and the function value at that point
Genetic cycle updating and stopping criteria.
is E t f xt , then the probability of the next point The end of last iteration, the values corresponding to
being at xt 1 depends on the difference in the function the string having minimum fitness value is identified.
values at these two points i.e. E E t 1 E t and
c) Particle Swarm Optimization [15]:
is calculated using the Boltzmann probability distribution Particle swarm optimization (PSO) is a computation
given by Eq. (17): technique developed by Eberhart and Kennedy based on
the analogy of swarm of birds and school of fish. PSO
E mimics the behaviour of individuals in a swarm to
P E t 1 min 1,exp (17) maximize the survival of the species.
kT
It is similar to other evolutionary computation
techniques in conducting searching for optima using an
If E 0 , this probability is one and the point xt 1 initial population of individuals ([16]-[20]).
is accepted. But if E 0 it implies that xt 1 is worse The individuals of this initial population are then
than that of xt . In this case according to Metropolis updated according to some kind of process such that they
are moved to a better solution area. In PSO system, each
algorithm, there is some finite probability of selecting the
individual adjusts its flying in a multi-dimensional search
point xt 1 even though it is worse than xt .
space according to its own flying experience and its
This probability depends on relative magnitude of E companions flying experience. Each individual is
and T . referred to as “particle” which represents a candidate
solution to the problem. Each particle is treated as a point
Algorithm:
in a D-dimensional space. The i th particle is represented
SA is a point by point method. The algorithm begins
with an initial point and a high temperature T . as X i xi1 ,xi 2 ,xi 3 ,.....xiD .
A second point is created at random in the vicinity of The best previous position (giving the best fitness
the initial point and the difference in the function value) of any particle is recorded and represented as
values f at these two points is calculated. Here in Pi Pi1 ,Pi 2 ,Pi 3 ...PiD . The index of the best particle
this work the difference in the function values is among all the particles in the population is represented
taken as f instead of E so as to correlate with by the symbol ' g' . The rate of position change (velocity)
performance index. for particle ‘ i ’ is represented as Vi Vi1 ,Vi 2 ,Vi 3 ....ViD .
If the second point has a smaller function value, the The particles are manipulated according to Eq. (18)
point is accepted otherwise the point is accepted with and (19):
a probability of (16). In order to do this, another point
is created at random in the neighborhood of the
current point and the Metropolis algorithm is used to
ViD ViD c1r1 PiD X iD c2 r2 PgD X iD (18)
accept or reject the point.
In order to simulate the thermal equilibrium at every X iD X iD ViD (19)
temperature, a number of points ' n' is tested at a
particular temperature before reducing the where c1 and c2 are positive constants and r1 and r2 are
temperature. The algorithm is terminated when
sufficiently small temperature and small enough uniformly distributed random numbers in the range (0,
change in function values are obtained. 1).
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
158
P. Marimuthu, B. Basavaraja, Subhransu Sekhar Dash
associated with the fittest solution it has achieved so of performance index obtained during various
far. optimization techniques.
The value of the objective function Pi is also tracked. It can be observed from Table I and Table II that the
The global version of the PSO keeps track of the performance of PSO is better than that of other
overall best value and its best location obtained so far optimization techniques in terms of Peak time, settling
by any particle in the population, which is called Pg . time and overshoot and also the performance index of the
system is very less as compared to the other optimization
The PSO at each step changes the velocity and techniques.
position of each particle towards its Pi and Pg . Fig. 7 shows the comparison of frequency deviation of
The particle swarm optimization then updates the both areas and tie line power error deviation with respect
particle coordinates based on (18) and (19) to all techniques for normal case. Fig. 8 shows the
These above two steps are repeated from population various generations of Gencos in thermal area during
to population until a stopping criterion terminates the normal case.
search producing the optimum gains. Fig. 9 shows the generations of Gencos in hydro area
during normal case.
TABLE I
COMPARISON OF PERFORMANCE OF OPTIMIZATION TECHNIQUES
Optimization Peak time Settling
S.No Integral Gain values Area Considered Overshoot (Hz)
Technique (s) Time (s)
Simulated ki1 1.35 Thermal area 2.73 0.00294557 7.04
1
Annealing ki 2 0.267 Hydro area 0.88 0.00590737 3.525
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
159
P. Marimuthu, B. Basavaraja, Subhransu Sekhar Dash
Fig. 10. Frequency and tie line power error deviations during
Fig. 8. Generation of Gencos of Thermal area during normal case
contract violation case
Fig. 9. Generation of Gencos of Hydro area Fig. 11. Generation of Gencos of Thermal area
during normal case during Contract violation case
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
160
P. Marimuthu, B. Basavaraja, Subhransu Sekhar Dash
Fig. 10 depicts the comparison of frequency deviation by Particle Swarm Optimization brings about better
of both areas and tie line power error deviation with response in terms of overshoot and settling time as
respect to all techniques for contract violation case. shown in the simulation.
Fig. 11 shows the various generations of Gencos in
thermal area during contract violation case. Fig. 12
shows the generations of Gencos in hydro area during
contract violation case. Fig. 13 and 14 depict the
performance index of the system for various techniques
during normal case and contract violation case.
TABLE II
COMPARISON OF PERFORMANCE INDEX VALUES FOR NORMAL CASE
S. No. Optimization Technique Performance Index
5
1 Simulated Annealing 3.241 10
Fig. 13. Comparison of Performance index values
5
2 Genetic Algorithm 2.15 10 during Normal case
Particle Swarm 5
3 1.91 10
Optimization
Appendix
(a) System data
T p1 , T p 2 =20s; K p1 , K p 2 =120Hz/p.u. Mw;
Pr1 , Pr 2 =1200Mw; Tt =0.3s; T g =0.08s, Tw =1s;
Tr 5s , T1 41.6s , T2 =0.513s; R1 , R2 = 2.4Hz/pu
Mw; T12 = 0.0866s; B1 , B2 0.4249p.u Mw/Hz;
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
161
P. Marimuthu, B. Basavaraja, Subhransu Sekhar Dash
[5] Jayant Kumar, Kah-Hoeng and Gerald Sheble, “AGC simulator Authors’ information
for price based operation part- 2”, IEEE Transactions on Power
Systems, Vol.12, no. 2, May1997, pp 533-538. P. Marimuthu was born in 1977. He received
[6] Bjorn H.Bakken and OvesGrande, “Automatic generation control his B.E. in Electrical and Electronics
in a deregulated environment”, IEEE Transactions on Power Engineering from Madras University, India in
Systems, vol.13, no.4, Nov1998,pp. 1401-1406. 1998, his M.E. in Power Systems from Anna
[7] V. Donde, M. A. Pai and I. A. Hiskens, “Simulation of bilateral University, Chennai, India in 2005. He is
contracts in an AGC system after restructuring”, Power systems currently working towards his Ph.D. in the field
engineering research center, Oct 2000. of Improvement of Dynamic performance of
[8] Rao, C. Srinivasa; Nagaraju, S. Siva; Raju, P. Sangameswara, A Multi Area Hydro Thermal System at Jawaharlal
modified genetic approach to hydrothermal system with thyristor Nehru Technological University (JNTU), Hyderabad, India. He is a
controlled phase shifter under open market system, (2007) member of the faculty in the Department of Electrical and Electronics
International Review of Electrical Engineering (IREE), 2 (4), pp. Engineering, Annapoorana Engineering College, Salem, India. His
507-514. research interest includes dynamic response of hydro thermal system,
[9] Marimuthu, P., Govindaraju, C., Load frequency control of optimal location and Power flow control with FACTS devices.
hydrothermal system under open market considering capacitive
energy storage, (2012) International Review on Modelling and Dr. B. Basavaraja was born in 1970. He
Simulations (IREMOS), 5 (5), pp. 2307-2313. obtained his B.Tech (EEE) degree from
[10] P.Marimuthu, S.S.Dash, B.Basavaraja “Performance Analysis of Gulbarga University, India, M.Tech from
CES and CES-SSSC in Load Frequency Control of Multi Area Karantaka University, India and PhD from NIT
System under Open Market Scenario” International Journal of Warangal, India. He is presently working as a
Distributed Energy Resources and Smart Grids, vol. 10 Number 1 Professor & Chairman (HOD), EEE, University
(2014) Page No. 1-13. of BDT Engineering College, Davanagere,
[11] Praghnesh Bhatt, S. P. Ghoshal, Ranjit Roy, “Optimized (Visvesvaraya Technological University
Automatic Generation Control by SSSC and TCPS in Belgaum) Davanagere, Karnataka, India. He is Executive Committee
Coordination with SMES for Two-Area Hydro-hydro Power Member of ISTE and Senior IEEE member. His area of interest
System”, Proceedings of the 2009 International Conference on includes power electronics and drives, High voltage Engineering and
Advances in Computing, Control, and Telecommunication EMTP applications.
Technologies, 2009, pp 474-480.
[12] “Dynamic models for steam and hydro turbines in power system Dr. Subhransu Sekhar Dash received the M.E
studies” IEEE committee report, IEEE Trans, PAS-92, 1973, pp degree in Electrical Engineering from UCE
1904-1915. Burla, Orissa, India and Ph.D degree in
[13] Kalyanmoy Deb, “Optimization for engineering design”, Prentice Electrical Engineering from Anna University in
Hall of India, 2000. 1996 and 2006 respectively. He is presently
[14] H. Shayeghi, H. A. Shayanfar, A. Jalili and M. Sivandian, “A working as Professor and Head of the
genetic algorithm based AGC of a restructured power system”, department of Electrical Engineering in SRM
International conference on Artificial Intelligence, Jun 2006, pp University, Chennai, India. His area of interest
237-242. includes Power Quality, Converters, Multilevel Inverters, Power
[15] L. Youssef Abdel-Magid and M. A. Abido, “AGC tuning of System Operation & Control, Stability and Intelligent controlling
interconnected reheat thermal systems with particle swarm Techniques.
optimization”, Proc of IEEE Conf. Electronics, Circuits and
Systems, 2003, pp 376-379.
[16] Shankar, T., Shanmugavel, S., Karthikeyan, A., Hybrid approach
for energy optimization in wireless sensor networks using PSO,
(2013) International Journal on Communications Antenna and
Propagation (IRECAP), 3 (4), pp. 221-226.
[17] Barzegari, A., Barforoushi, T., Asgharpour, H., Lesan, S., A new
algorithm based on particle swarm optimization for solving power
economic dispatch considering valve-point effects and emission
constraints, (2011) International Review on Modelling and
Simulations (IREMOS), 4 (3), pp. 1303-1311.
[18] Sakthivel, S., Mary, D., Voltage stability limit improvement by
static VAR compensator (SVC) under line outage contingencies
through particle swarm optimization algorithm, (2011)
International Review on Modelling and Simulations (IREMOS), 4
(2), pp. 766-771.
[19] Paramasivam, B., Chidambaram, I.A., Design of a load-frequency
controller using craziness based PSO for an interconnected power
system with SSSC and RFB, (2012) International Review of
Automatic Control (IREACO), 5 (2), pp. 102-112.
[20] Ibrahim, H.E.A., Elnady, M.A., A comparative study of PID,
fuzzy, fuzzy-PID, PSO-PID, PSO-fuzzy, and PSO-fuzzy-PID
controllers for speed control of DC motor drive, (2013)
International Review of Automatic Control (IREACO), 6 (4), pp.
393-403.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
162
International Review of Electrical Engineering (I.R.E.E.), Vol. 10, N. 1
ISSN 1827- 6660 January – February 2015
Keywords: Autonomous, Flight Envelopes, Overhead Power Lines, Robotic, Supervisory Control
163
S. Antunes, K. Bousson, J. Mota
the system model error and measurement noise area and the surrounding environment characteristics.
covariance matrices are just set subjectively. For that The proposed obstacle avoidance methods for
reason, accurate attitude estimation based on model overhead line inspection are far from being optimal since
uncertainty approach has been proposed by Bousson and they do not deal with minimizing the distance between
colleagues [12] for flight control of unmanned aerial the flight tracks during the avoidance process to the
vehicles. The Kalman filtering method is explained planned waypoint trajectory that was initially scheduled.
hereafter. Let a continuous-time system be described by Recently advances in optimal collision avoidance
the following state-space equations: methods and optimal 4D trajectory generation conducted
by Bousson and colleagues [16]-[18] offer interesting
x Ax Bu w t insight that could improve the collision avoidance task
(1) for airborne inspection of power lines.
y Cx v t The critical problems in airborne overhead line
inspection are related not only to acquiring accurate
where x is the state vector, u the control vector, and y video imaging for efficient defect detection on overhead
the output vector. Vectors w t and v t are lines and vegetation assessment but also, and mainly, to
flight safety.
respectively the system model uncertainty dynamics and Indeed, due to the low flight altitude and speed of the
the observation noise. Let Q E wwT and R vvT aircraft that are required for precise operations, most of
the flight will be held outside regular safety
be defined as the covariance matrices of w t and v t , configurations for aircraft operations on minimum
respectively. Then, assuming the system to be distance to ground and obstacles, altitude above ground
controllable and observable, the Kalman filter model is to instant air speed ratio, time for decision on
given as follows: contingencies, especially when hovering to assess the
The state estimate model: line conditions or when tracking the lines, span after
span. Current research activities about airborne
ˆx Axˆ Bu L y Cxˆ ; x t0 ˆx0 (2) inspection of power lines have not dealt yet with these
serious flight safety problems. Safe fight operations are
where gain matrix L is defined as: achievable by supervising the aircraft maneuvers and
monitoring them within the flight envelope during the
inspection.
L PC T R 1 (3)
Another aspect that has not got sufficient attention in
airborne power line inspection is real-time telemetry, that
where P is the solution of the following Riccati matrix
is, sending inspection data as well as flight characteristics
differential equation:
(state, parameters, navigation trajectory, ...) to a distant
station to help improve monitoring of the inspection
P AP PAT Q PC T R 1CP (4) process while the aircraft is still in the air, thus allowing
the inspection crew on the ground to interact with the
Similarly, in the case of discrete-time system given vehicle (and airborne crew, if present) when more visual
by: information of the inspected item needs to be performed.
xk 1 uk uk wk Besides, corrosion detection shall be based on mixed
(5) approach built on pattern recognition methods and
yk Cxk vk physical principles [19]. The proposed project will tackle
these issues by providing a better understanding of safe
with the estimative: supervisory control and monitoring strategies for
unmanned aerial vehicle operations with respect to
ˆxk 1 ˆxk u k Lk yk Cxˆ k ; x t0 ˆx0 (6)
stringent flight modes within tight flight operational
envelopes during power line inspection.
The gain was given by:
1 II. Overhead Line Inspections
Lk Pk C T R CPk C T (7)
The constant evolution of technology, and need for the
with Riccati solution as follow: operator to perform missions safely, suggests the use of
autonomous platforms replacing humans in some of their
1 works in order to increase safety of involved platforms,
Pk 1 Q Pk T Pk C T R CPk C T CPk T (8) protect human lives, extend time for carrying out
missions and acquire data, and reduce the risk factors
Obstacle detection and avoidance methods based on such as human fatigue when exposed to danger work.
computer vision are dealt with in [13]-[15] with some Albatroz Engineering designed a system to inspect
relative success, depending on the illumination of the overhead lines using different aerial and ground vehicles.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
164
S. Antunes, K. Bousson, J. Mota
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
165
S. Antunes, K. Bousson, J. Mota
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
166
S. Antunes, K. Bousson, J. Mota
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
167
S. Antunes, K. Bousson, J. Mota
In the case of the intended application with the use of Enter Exit
unmanned aerial vehicles, the aim is to make flight
inspection respecting the safety limits taking into account Inspection System
flight procedures specific.
Although the payload applied to aircraft provided it is
within the maximum load will not be necessary to check
the structural limits of the aircraft, it should be noted is
not a goal designing an aircraft, but to verify Control
qualitatively and quantitatively how we can approach the
target to effect inspecting, maintaining a certain speed or
even effecting a ploy to 'hover' without exceeding the
limits and safety margins allowing the pilot to make Surveillance
flight maneuver not to hit the object analysis.
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
168
S. Antunes, K. Bousson, J. Mota
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
169
S. Antunes, K. Bousson, J. Mota
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
170
S. Antunes, K. Bousson, J. Mota
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved International Review of Electrical Engineering, Vol. 10, N. 1
171
1827-6679(201501/02)10:1;1-K
Copyright © 2015 Praise Worthy Prize S.r.l. - All rights reserved