IMPROVED COST OF ENERGY COMPARISON OF

PERMANENT MAGNET GENERATORS FOR LARGE

OFFSHORE WIND TURBINES
Kerri Hart Alasdair McDonald Henk Polinder
University of Strathclyde University of Strathclyde Delft University of Technology
[email protected] [email protected] [email protected]

Edward Corr James Carroll

University of Strathclyde University of Strathclyde
[email protected] [email protected]

Abstract: developers such as installation and

scheduled/unscheduled maintenance. Weather
This paper investigates geared and direct-drive windows can be unpredictable and often too short
permanent magnet generators for a typical to carry out significant repairs and so developing
offshore wind turbine, providing a detailed reliable and efficient wind turbines is of extreme
comparison of various wind turbine drivetrain importance. Large wind turbines (>5 MW) are now
configurations in order to minimise the Cost of being manufactured to allow a fewer number of
Energy. The permanent magnet generator turbines to be installed in an offshore site whilst
topologies considered include a direct-drive maintaining high energy yield. By reducing the
machine and single stage, two-stage or three- number of installed turbines the service and repair
stage gearbox driven generators. The cost of requirements can be minimized.
energy calculations are based on initial capital
costs, the costs of manufacture, installation, Over the past few decades, wind turbine
operations and maintenance, with particular focus manufacturers have been exploring various drive
on improved calculations of the annual energy train topologies ranging from multistage gearbox
yield with better availability estimations and and induction generators to gearless direct-drive
gearbox loss modelling. systems. With high emphasis on turbine reliability
for modern offshore developments, some
Keywords: Cost of energy, permanent magnet manufacturers are beginning to concentrate on
generator, direct-drive, gearbox, wind turbine. using direct-drive generators which do not have a
gearbox (eliminating gearbox related failures).
Direct-drive technology might offer reduced
1 Introduction maintenance cost and complexity and improved
reliability [3]. However, these slow speed
This paper builds on the work of Polinder et al. [1] generators need to be large, able to deliver high
and Bywaters et al. [2] but with an emphasis on a torque and robust enough to cope with large forces
typical 6MW offshore wind turbine. This paper that might be avoided by using a gearbox.
considers the four following generator topologies: Therefore, direct-drive machines tend to be large,
the direct-drive permanent magnet generator heavy and are expensive.
(DDPMG), the permanent magnet generator with
single stage gearbox (PMG1G), the permanent With an aim to further enhance performance and
magnet generator with two-stage gearbox reliability, manufacturers are also implementing
(PMG2G), and the permanent magnet generator permanent magnet (PM) systems into their
with three-stage gearbox (PMG3G). turbines to eliminate excitation losses. Permanent
magnet machines do not require additional power
Offshore wind farm development has rapidly supply for the magnet field excitation and have
increased in the past few years. The issues and higher efficiency and reliability [3] compared with
limitations present in onshore wind farm electrically excited machines. Sizing and cost
development, such as turbine size and noise become significant issues with direct-drive PM
production, are much less prevalent offshore, and configurations as generators may require very
combined with excellent wind resources offshore large diameters and so can be heavy and
wind farms are becoming more attractive to expensive which in turn can create transport and
developers on a global scale. However, the harsh installation issues. As a result, turbine
sea environment poses several issues for manufacturers have also been considering hybrid
systems consisting of a permanent magnet 2 Modelling Methodology
generator and single-stage or 2-stage gearboxes.
The gearbox and generator can either be separate The design of the wind turbine and the generator
or integrated together, for example the Multibrid configurations are based on approximations and
systems – such as the AREVA M5000. Both values found in Polinder’s paper [1]. The generator
approaches represent a compromise between high dimensions of the 3 MW wind turbine that form the
speed geared systems and direct-drive systems. basis of that study are scaled by the generator
Medium speed solutions require a smaller torque to maintain consistent comparisons for the
generator than a direct-drive generator. 6 MW turbine considered in this paper. The geared
generator concepts have their gear ratio design
In this paper the authors limit analysis and based on current manufacturers design and some
discussion to drivetrains with permanent magnet limited optimisation. The costs of the generator
generators – which are currently the preferred configurations are based on the mass of materials
solution for a majority of manufacturers of offshore used in each generator and material cost in Table
wind turbines. The generator systems considered 1. Other costs such as construction and the costs
in this paper are as follows: of the converter are scaled with an inflation rate of
2.4% p.a. [4] to be in accordance with the present
 DDPMG – Direct-drive permanent magnet day value.
generator
 PMG1G – Permanent magnet generator,
single-stage gearbox
Wind Turbine Characteristics
 PMG2G – Permanent magnet generator,
Rated Grid Power (MW) 6
two-stage gearbox
Rotor Diameter (m) 140
 PMG3G – Permanent magnet generator, Rated Wind Speed (m/s) 11
three-stage gearbox Rated Speed (rpm) 12
Optimum Tip Speed Ratio 8
Whilst being cognisant of promising alternatives to Maximum Aerodynamic Efficiency (%) 48
the gearboxes (hydraulic and electromechanical Mass density of air (kg/m3) 1.225
gearboxes) and PM generators (DFIGs, brushless Generator Material Characteristics
DFIGs and HTS generators), the authors want to Slot filling factor ksfil 0.6
address the question ―which PM generator Remnant flux density of magnets Brm (T) 1.2
Recoil permeability of the magnets μrm 1.06
drivetrain delivers the lowest cost of energy
Resistivity of copper at 120οC ρCu (μΩm) 0.025
offshore?‖ The variation in turbine designs Eddy-current losses in laminations at 1.5 T, 50 Hz
suggests that different manufacturers have come 0.5
PFe0h (W/kg)
to different conclusions. Hysterisis losses in laminations at 1.5 T and 50 Hz
2
PFe0h (W/kg)
The contribution of this paper is that it compares Loss Modelling
different permanent magnet generators on a cost Maximum losses in a 6 MW VSI Pconvm (kW) 180
of energy basis. The paper begins with a section Cost Modelling
on the modelling of the wind turbine, gearbox, Single-stage gearbox (ratio 8) cost (kEuro) 672
converter and the generator. The details of the four Two-stage gearbox (ratio 40) cost (kEuro) 1170
different generator topologies are then described Three-stage gearbox (ratio 100) cost (kEuro) 1330
and their simulated performances are analysed Power electronics cost (Euro/kW) 40
Laminations cost (Euro/kg) 3
which includes a sensitivity analysis. The paper
Copper cost (Euro/kg) 15
concludes with a comparison of the four generator Permanent magnet cost (Euro/kg) 48
concepts and which configuration may offer the Rest of wind turbine cost (Euro) 6100
most economic cost of energy.
Table 1: Modelling Characteristics.
A baseline scenario is presented with typical
drivetrain losses, costs and ratings. Using the 2.1 Wind Turbine Model
same designs, further scenarios are investigated.
In two scenarios the gearbox or the generator is The comparison of the four generator systems was
replaced once in its lifetime for each of the achieved through the simulation of a theoretical 6
drivetrains, and in another the cost of permanent MW wind turbine with a rated speed of 12 rpm.
magnets increases. These represent credible risks The aerodynamic power of a wind turbine is given
that drivetrain designers must consider.
by the following equation:

(1)
where is the mass density of air, r is the rotor
radius of the wind turbine, vw is the wind speed, (5)
and Cp is the power coefficient (or aerodynamic
Cost functions for two-stage gearboxes are more
efficiency) which is a function of tip speed ratio
difficult to find. In this case we assume an average
and pitch angle . The Weibull distribution
of equations (4) and (5); this gives equation (6). In
that describes the variation in wind speed at a
this study the gear ratio is chosen to be 40.
given site over a year is given in equation (2)
below: (6)
( ) [ ( ) ] (2) In previous studies (e.g. [1], [6]) gearbox losses
have been approximated based on an assumption
where k is the shape parameter, C is the scale that 1% of viscous losses per gearbox stage are
parameter, and v is the wind speed. reasonable. Reference [1] used x = 3% for a 3-
stage gearbox and x=1.5% for a single-stage
The annual energy production (AEP) is calculated gearbox and scaled the loss using a ratio of actual
based on the Weibull probability distribution: rotational speed (n) to rated rotational speed
(nrated). Prated is the wind turbine power:
∫ (3)
(7)
where vi is the cut-in wind speed and vc is the cut-
out wind speed if the turbine.
The losses in an example 3-stage gearbox
The power curve and Weibull distribution of the topology [7] were calculated according to equation
turbine are shown in Figure 1. The cut-in wind (7). These are shown by the ―Scaled 3% loss
model‖ curve in Figure 2 for a 2.5MW turbine. A
speed is 4 m/s and the cut-out wind speed is 25
more sophisticated approach was then used to
m/s with a rated power of 6 MW being attained at a
calculate the losses. ISO/TR 14179-1:2001
wind speed of 11 m/s. The parameters for the
standards [8] specify likely losses for cooling
Weibull distribution are based on a location in the requirements and where applied to the gear and
North Sea with a mean wind speed of 9.8 m/s, a bearing data from [6]. The losses from this method
scale parameter of 10.8 and a shape parameter of are given in Figure 2 by the ―ISO method‖ curve. It
2.32. shows that equation (7) can overestimate gearbox
losses over the whole wind speed range and
0.1 particularly at low-medium wind speeds.
6
Power (MW)

Probability

0.08
4 0.06
2 0.04
0.02
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Wind Speed (m/s) Wind Speed (m/s)

Figure 1. Power curve and Weibull distribution

as a function of wind speed.

2.2 Gearbox Modelling

The cost of gearboxes is difficult to estimate given Figure 2. Comparison of gearbox loss models for a
the variety in topologies and manufacturers. 2.5MW wind turbine
However Fingersh et al. give some useful
expressions as a starting point [5].

Reference [5] gives an estimate of the cost (in €)

of the single-stage gearbox following equation (4).
In this case we use a gear ratio of 8.

(4)

The ratio of the three-stage gearbox is chosen to

be 100 and an estimate of the cost of the gearbox
is given by [5]:
Figure 3: Gearbox efficiency curves used in
this study
The data from [6] was used to calculate losses The flux density directly above a magnet in the air
associated with the first, second and third stages gap of a permanent magnet machine is calculated
of the gearbox. These losses were then scaled to as:
the required gearbox sizes for this study. The
gearbox efficiency curves for the chosen designs ̂ ( ) (11)
are presented in Figure 3.
where Brm is the remanent flux density of the
2.3 Converter Modelling magnets (1.2T), bp is the magnet width and τp is
the pole pitch.
In this study all of the drivetrains require a fully
rated converter as permanent magnet generators The no-load voltage induced by the flux density in
are used. a stator winding is given by:

Back-to-back power converters can be used as the √ ̂ (12)

interface between the turbine generator and the
where ωm is the mechanical angular speed of the
grid by implementing PWM (pulse width
rotor. The copper losses in the generator are
modulation) to ensure a steady sine-wave output 2
calculated from the currents and resistances (I R
from the variable speed operation of the turbine.
losses). The specific iron losses are given by:
The power converter consists of a series of IGBTs
that make up a generator side converter, a grid ̂
side inverter, and a DC-link capacitor. The losses ( )( )
̂
in the power converter are modeled using the (13)
following equation: ̂
( ) ( )
̂
(
where fe is the frequency of the field in the iron,
(8) is the hysteresis loss per unit mass at the
) given angular frequency f0 and flux density B0, and
is the eddy current loss per unit mass.
where is the dissipation in the converter at
rated power (3% of the rated power of the 2.5 Cost of Energy
converter), Is is the generator side converter
current, Ism is the maximum generator side The cost of energy (COE) [2] per MWh is the
converter current, Ig is the grid side inverter overall outcome of this study and is calculated
current, Igm is the maximum grid side inverter using the following equation:
current [1].
(14)
2.4 Generator Modelling
where FCR is the Fixed Charge Rate, ICC is the
Equivalent circuit models are used to compare the Initial Capital Cost of the turbine, AOM is the
generators are described in [1]. The magnetized Annual Operation and Maintenance and AEP is
inductance of an AC machine is given by: the Annual Energy Production.

(9) In this case ICC is made up of a fixed part which

represents the rest of the wind turbine (common to
where ls is the stack length in axial direction, rs is all of the designs) and a part which equals the cost
the stator radius, Ns is the number of turns of the of the generator, any gearbox and the power
phase winding, kw is the winding factor, p is the converter.
number of pole pairs, and geff is the effective air
gap. The effective air gap is given by: AOM is calculated according to [5] which uses a
cost/MWh energy produced ratio for calculating
( ) (10) operation and maintenance costs. It ignores the
effect of drivetrain choice of maintenance cost –
where ksat is a factor representing the reluctance of unfortunately there is a lack of data that
the iron in the magnetic circuit, kCs is the Carter distinguishes operation and maintenance costs of
factor for the stator slots, g is the mechanical air different types of offshore wind turbine drivetrains.
gap, µrm is the relative recoil permeability of the
magnets, and lm is the magnet length in the AEP is calculated using the methods described in
direction of the magnetization. Section 2 as well as availability data given in Table
2. These availability figures are based on failure Results from the MATLAB model for the DDPMG
rates that are synthesised from various reliability are shown in Figure 4. The results shown include
studies and onshore mean time to repair modified voltage and current levels, the wind turbine power
to include extra downtime for offshore work and curve, the system efficiency and losses. The
delays. For more details the reader is referred to system efficiency takes into account bearing and
the paper by Carroll [9]. cable losses as well as generator and converter
losses. In this case the generator diameter was
3 Results restricted to 7m. The significant losses in the
system result from high copper losses that account
The results for all four designs are presented in for over half of the total losses. This is due to the
Table 2. The following sections details the loss requirement that in order to produce a high torque
mechanisms and costs for each of the designs. there is a large number of coils. Availability is high:
even though there are increased winding failures
DD PMG PMG PMG in the direct-drive generator, there is no downtime
PMG 1G 2G 3G
Generator Specifications due to a gearbox.
Generator speed (rpm) 12 96 480 1200
Gearbox ratio - 1:8 1:40 1:100
Stator radius rs (m) 3.5 2.5 0.7 0.5
Stack length ls (m) 1.5 0.4 0.8 0.9
Number of pole pairs p 100 54 10 3
Air gap g (mm) 7 5 1.4 1
Stator slot width bss (mm) 17 21 19 33
Stator tooth width bst (mm) 20 27 23 40
Stator slot height hss (mm) 80 80 80 80
Stator yoke height hsy (mm) 40 40 40 40
Rotor yoke height hry (mm) 40 40 40 40
Magnet height lm (mm) 15 15 15 15
Magnet width bp (mm) 87.5 116 178 417
Generator Parameters
Main inductance Lm (mH) 8.3 1.9 3.7 8.3
Stator leakage inductance
Lsσ (mH) 10.4 1.3 0.9 0.4
Stator resistance Rs (mΩ) 181 32.5 20.9 12.1
Generator Material Weight (ton)
Iron 30.6 6.4 3.3 2.8
Copper 6.6 2.0 1.0 1.1
PM 2.9 0.6 0.3 0.3
Total 40.1 9.0 4.6 4.2 Figure 4: DDPMG Drivetrain Operation &
Cost (kEuro) Performance
Generator active material 330 77 39 38
Generator construction 436 115 24 10
Gearbox - 672 1170 1330 3.2 PMG1G
Generator system cost 767 864 1240 1380
Converter 283 283 283 283
Other wind turbine parts 6000 6000 6000 6000 This drivetrain a similar but slightly higher Cost of
Total cost of wind turbine 7050 7250 7520 7660 Energy than the direct-drive machine. Although it
Annual Energy
has a lower rating than the turbine discussed here,
Copper losses (MWh) 1120 201 135 79
Iron losses (MWh) 151 242 243 529 the AREVA M5000 would be a good exemplar of
Converter losses (MWh) 880 861 868 894 the single-stage PM generator drivetrain. Results
Gearbox losses (MWh) 0 314 779 879 are shown in Figure 5 and Table 2.
Availability (%) 93.4 93.0 92.8 92.6
Total losses (MWh) 2150 1620 2030 2380
Energy yield (GWh) 28.0 28.6 28.3 28.2 Because of its smaller torque rating the electrical
Cost of Energy machine is smaller and cheaper (in terms of
ICC (kEuro) 21400 22000 22800 23200 materials and construction) than the direct-drive
AOM (kEuro) 628 641 635 632
FCR 0.116 0.116 0.116 0.116 generator. The addition of the gearbox does make
COE (Euro/MWh) 110.7 111.3 115.6 117.7 it more expensive (than the direct drive machine)
from the view point of capital costs.
Table 2: Design details, costs and performance
of 4 drivetrains
The real benefit of a PMG driven by a single-stage
gearbox is the low losses with copper losses, iron
3.1 DDPMG losses and gearbox losses all being fairly
balanced.
Overall this drivetrain gave the lowest Cost of
Energy in the baseline study. The Siemens SWT- Although there is a gearbox – which means that
6.0 150 and Alstom Haliade turbines – although there are failures and downtime over and above
very different machines – fit into this category. the direct-drive machine – there are less electrical
failures because the generator is smaller. Offshore
availability is very similar to the direct-drive However, due to having a two-stage gearbox, the
machine. gearbox losses become more significant and the
gearbox itself is larger and more expensive.
3.4 PMG3G
The PMG driven by a 3-stage gearbox has not yet
been a popular choice in the offshore wind market,
although the leading generator manufacturers
such as ABB and the Switch have high speed
PMGs in the 5-7MW range. As one might expect,
the generator is very compact, cheap and efficient.
Unfortunately the increased losses, cost and
downtime due to the additional gearbox stage give
rise to a higher Cost of Energy than all of the other
drivetrains. Results from the MATLAB model for
the PMG with a 3-stage gearbox are shown in
Figure 7 and Table 2. It might be attractive should
the cost of permanent magnets become very high
Figure 5: PMG1G Drivetrain Operation & or if owners believe that they will have to replace
Performance many generators in their fleet.

Figure 6: PMG2G Drivetrain Operation & Figure 7: PMG3G Drivetrain Operation &
Performance Performance

3.3 PMG2G 4 Sensitivity Analysis

The Gamesa G128-5.0 MW and Samsung S7.0- Once designed and in production, the Cost of
171 turbines adopt a 2-stage gearbox with a Energy of an offshore wind turbine is somewhat
permanent magnet generator. Results are shown dependent on future events such as material price
in Figure 6 and Table 2. changes and sub-assembly failure rates. In this
section the model is used to look at the sensitivity
Here the Cost of Energy is higher than the direct- to some of these factors.
drive and PMG1G. Overall losses and costs are
higher than the drivetrains with slower generators; DD PMG PMG PMG
PMG 1G 2G 3G
availability is marginally worse because of the
COE (Euro/MWh) 110.7 111.3 115.6 117.7
added failures in the gearbox when a second
COE: 1 gearbox
stage is added. replacement (Euro/MWh)
110.7 114.4 120.9 123.8

COE: 1 generator
114.0 112.2 115.9 117.8
The generator size is considerably smaller than replacement (Euro/MWh)
the single-stage gearbox design and so has Table 3: Sensitivity analysis for different
reduced losses in the stator. This lightweight scenarios
generator would also be advantageous to
developers during installation procedures.
4.1 Gearbox and generator 5 Discussion and Conclusions
replacement scenarios
This study provides a comparison between four
different drivetrain configurations using permanent
There have been some wind farms where there
magnet generators. The drivetrains were modelled
have been a significant number of gearbox
to assess which drivetrain configuration offers the
replacements required. What happens to the cost
lowest Cost of Energy solution for a large offshore
of energy if there is one gearbox replacement per
wind turbine.
turbine over the lifetime of the wind farm? This can
be modelled by modifying equation (14) and
A direct-drive option can deliver the lowest Cost of
adding a replacement cost for one gearbox
Energy. This is shown in both the baseline study
replacement or one generator replacement. Table
and when a gearbox replacement is required (for
3 shows the change in cost of energy (from the
the other candidate designs). Permanent magnet
baseline study in section 3). The increase in Cost
generators have a limited track record in the wind
of Energy for the geared drivetrains demonstrates
industry (particularly offshore) and so the scenario
gives a clearer advantage for the direct-drive
of a generator replacement – once during the
machines.
turbine’s lifetime – is not unreasonable. Here the
direct-drive generator is second best. Even under
On the other hand the geared drivetrains are less
significant permanent magnet cost increases (i.e.
sensitive to the cost of having to replace a
×2.5) this drivetrain fairs well, though once the
generator once in the wind turbine’s life. Under this
magnet price increases above about €60/kg it is
scenario it is the single-stage design which has the
inferior to the single-stage gearbox machine.
lowest Cost of Energy.
Some potential disadvantages of the direct-drive
4.2 Permanent magnet cost increase generator have not been captured in this study:
scenario any increased wind turbine costs if the tower,
The last scenario in this study looks at the foundation and installation costs increase because
sensitivity of Cost of Energy to changes in if increased top head mass is large; also changes
permanent magnet prices. Future changes in in operation and maintenance costs which depend
magnet costs mainly affect a portion of the on the drivetrain. The availability has been based
generator part of the initial capital cost (ICC) on onshore failure rate data in the absence of real
component in equation (14). The baseline cost of operating data of offshore wind turbines with
the magnets is €48/kg; here the same baseline direct-drive PMGs.
designs but with a scenario of €120/kg are also
tested. Figure 8 shows these points. If all other In this study the diameter of the direct-drive
factors are constant then the Cost of Energy generator was limited to 7m – this was assumed to
changes linearly. The direct-drive design has the be pragmatic in terms of transportation – and
highest gradient, reflecting the high magnet based on an air-cooled design. The optimisation
content in lower speed, higher torque generators. consistently delivered designs at this limit which
One can interpret Figure 8 as showing that a future suggests that a large diameter machine would
magnet cost change from €48/kg to €60/kg would deliver an even lower Cost of Energy (by
make the permanent magnet generator with a producing higher efficiency with the same amount
single-stage gearbox more attractive than the of expensive active material). This is an
direct-drive option. opportunity for offshore wind turbines where
transportation restrictions are not as strict as
onshore.

The PMG1G performs well in terms of both energy

yield and COE. The single stage gearbox concept
outperforms the other gearbox designs, reinforcing
the benefits of the ―Multibrid‖ design. In the
scenario with one gearbox replacement, there was
a slightly bigger difference between the COE for
the DDPMG and the PMG1G. For scenarios with a
generator replacement or when the magnet cost
increases by a reasonable amount this drivetrain
delivers the lowest Cost of Energy.
Figure 8: Sensitivity to change in permanent
magnet price. This study did not define the amount of integration
of generator and gearbox; it is possible that some
highly integrated designs might lead to higher Trans. Energy Convers., vol. 24, no. 1, pp.
maintenance and replacement costs. 82–92, Mar. 2009.

The higher speed generator drivetrains faired [4] Trading Economics, Euro Area Inflation
much worse under the baseline and other Rate. [Online] [Cited 20/7/12]
scenarios. Gearbox cost reductions below the http://www.tradingeconomics.com/euro-
assumed values predicted by equations (5) and (6) area/inflation-cpi.
may help reduce Cost of Energy to more
competitive levels under the combined scenarios [5] L. Fingersh, M. Hand, and A. Laxson,
of higher magnet costs and a generator
―Wind turbine design cost and scaling
replacements.
model,‖ no. December, 2006.

Acknowledgments [6] J. Cotrell, "A preliminary evaluation of a

multiple-generator drive train configuration
This work has been funded by the EPSRC, project for wind turbines." In 21st American Society
reference number EP/G037728/1. of Mechanical Engineers (ASME) Wind
Energy Symposium, 2002.
References
[7] S. Sheng and C. Broomfield, ―Wind Turbine
Gearbox Condition Monitoring Round
[1] H. Polinder, F. F. A. Van Der Pijl, G. De
Robin Study – Vibration Analysis," NREL,
Vilder, and P. J. Tavner, ―Comparison of
Golden, Colorado, Report no. NREL/TP-
Direct-Drive and Geared Generator
5000-54530, 2012.
Concepts for Wind Turbines,‖ IEEE Trans.
on Energy Convers., vol. 21, no. 3, pp. 725-
733, 2006. [8] Anon, ―Gears — Thermal capacity —Rating
gear drives with thermal equilibrium at 95°C
sump temperature," ISO/TR 14179-1:2001,
[2] G. Bywaters, V. John, J. Lynch, P. Mattila,
2001.
G. Norton, J. Stowell, and D. Hablanian,
"Northern Power Systems WindPACT drive
train alternative design study report." [9] J. Carroll, A. McDonald, ―Drivetrain
NREL, Golden, Colorado, Report no. Availability in Offshore Wind Turbines,"
NREL/SR-500-35524, 2004. Abstract submitted to EWEA 2014,
Barcelona 2014.
[3] H. Li, Z. Chen and H. Polinder,
―Optimization of Multibrid Permanent-
Magnet Wind Generator Systems,‖ IEEE