Pereira 2011 - Intro Engineering Final Project
Pereira 2011 - Intro Engineering Final Project
Jo
ao Pedro Pereira
[email protected]
Contents
1 Introduction
2 Physical limits to wind energy
2.1 Kinetic energy of wind . . . .
2.2 Wind speed variation . . . . .
2.2.1 Across space . . . . .
2.2.2 Through time . . . . .
.
.
.
.
2
2
3
4
5
.
.
.
.
.
.
.
.
.
5
5
7
7
7
9
10
10
10
12
13
13
15
17
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Introduction
2
2.1
(1)
This has units kg.m2 /s2 = (kg.m/s2 ).m = N.m = J. Consider that the
mass of air is passing through a vertical area A (which will be a circle for a
2
vt
typical horizontal axis wind turbine) and that the length of the mass of air
is L, as in gure 1.
Since mass equals volume times density, we have m = A L , where
is the density of air. At velocity v, during time t, we have L = v t. Thus,
m = A vt
(2)
1
1
KE = (A vt )v 2 = A t v 3
2
2
(3)
Replacing in (1),
Since power equals energy per unit of time, the power of wind is
1
P = KE/t = Av 3
2
(4)
kg
2 m
2 2
This has units m
3 .m . s3 = (kg.m /s )/s = J/s = W . Interestingly, power
depends on the cube of wind velocity.
2.2
Wind is second-hand solar energy. The Sun radiation heats the Earths surface, which in turn heats the surrounding air. Dierent areas (sea, mountains,
deserts, forests) absorb radiation dierently, so dierences in temperature
arise. The hotter air expands and rises, creating a pressure vacuum into
which cooler air rushes, which results in winds. The Earths movement then
generates variability in these relations through time. The nal result is that
the wind speed varies across space and over time.
2.2.1
Across space
Figure 2 shows the average speed across Portugal, estimated for a height of
60 m. Most of the territory shows an average speed around 5 or 6 m/s.
Figure 2: Wind speed in Portugal at height of 60m (Source: Esteves (2004))
Through time
At each location, the wind speed varies through time. According to Andrews
and Jelley (2007), the wind speed distribution is often well described by the
Rayleigh distribution, which has density
f (v) =
[
]
2v
exp (v/c)2
2
c
(5)
where c = 2
v / and v is the average wind speed. Figure 3 shows this
function for v = 6 m/s.
Figure 3: Rayleigh pdf for a mean wind speed of 6 m/s
0.14
0.12
Density, f(v)
0.1
0.08
0.06
0.04
0.02
3
3.1
10
15
20
Wind speed (m/s)
25
30
A windmill cannot absorb all kinetic energy in the wind. To do that, it would
have to fully stop the wind coming through, but then this slowed-down air
5
would get in the way. The windmill has to leave the air with some kinetic
energy. Figure 4 shows how the air slows down and sprays out as it passes
through a windmill.
Figure 4: Mass of air passing through a windmill (Source: MacKay (2009))
(6)
The physicist Albert Betz showed in 1919 that the maximum fraction of
the incoming energy that can be extracted by a windmill is 59%, that is,
max Cp = 0.59.
Proof. (Betz limit intuitive proof) Let v1 and v2 denote, respectively, the
wind velocity before and after the rotor. From equation (4), the full power
of the incoming wind is
1
P1 = Av13
2
Let KE1 and KE2 denote, respectively, the kinetic energy of the wind before
and after the rotor. By conservation of energy, the energy extracted by the
rotor, Er , must equal
1
Er = KE1 KE2 = m(v12 v22 )
2
where we used (1). From (2), m = A vr t , where vr is the wind velocity
through the rotor. Assume that this speed is an average of the speeds before
2
(in the more formal proof of Betz this a
and after the turbine, vr = v1 +v
2
result, not an assumption). The power extracted by the rotor is thus
)
(
1
1
v1 + v2
Pr := Er /t =
(v12 v22 ) = A(v1 + v2 )(v12 v22 )
A
2
2
4
6
The ratio between the power extracted by rotor and full incoming power is
(
( )2 ) (
)
1
v2
Pr
1 (v1 + v2 )(v12 v22 )
v2
Cp =
=
=
1
1+
P1
2
v13
2
v1
v1
Dene x := v2 /v1 and write the eciency as a function of x,
Cp =
)
1(
1 x2 (1 + x)
2
By maximizing this function, we nd arg maxx Cp = 1/3, that is, the maximum eciency is achieved when the departing wind speed is 1/3 of the
arriving wind speed, v2 = (1/3)v1 , which means that the speed at the rotor
is vr = (2/3)v1 . Replacing the optimum x = 1/3 above, we get the maximum
power coecient:
max Cp = 16/27 = 0.59
x
3.2
3.2.1
Power curve
Interestingly, the power output of a turbine does not increase monotonically with wind speed. Instead, the actual power output increases from the
minimum cut-in wind speed until it reaches the rated power at the rated
wind speed. After this rated wind speed, the power output stays constant at
the rated value and eventually drops to zero if the wind speed exceeds the
maximum safe operating speed (cut-out speed).
Figure 5 shows the power curve for the Vestas V112 turbine. Note that
this is a variable-speed variable-pitch turbine, so the power curve already
reects any possible optimal adjustments for each wind speed.
However, looking only at the power curve may be misleading. For example, one might conclude that the Vestas V112 turbine is only indicated for
7
Rated
Power
Turbine
Enercon E53
Vestas V90
Enercon E82
Siemens SWT 113
Vestas V112
Siemens SWT 120
Enercon E126
(kW)
Cut-in
wind
speed
(m/s)
Rated
wind
speed
(m/s)
Cut-out
wind
speed
(m/s)
800
000
000
300
075
600
500
2
4
2
3
3
4
3
13.0
13.5
13.0
12.5
13.0
12.5
16.0
28
25
28
25
25
25
28
2
2
2
3
3
7
Rotor Hub
diam- height
eter
(m) (m)
52.9
90
82
113
112
120
127
60
80
78
100
84
90
135
sites with wind speeds close to 13 m/s. The same applies to all other turbines
in table 1. This is odd as there are very few sites with average speeds as high
as 13 m/s... In reality, these turbines are actually optimized for much lower
average wind speeds, as discussed in the following section.
3.2.3
Power coecient
What really determines the performance of a given turbine is its power coecient (Cp , as dened in equation (6)). Figure 6 shows both the power
curve and the power coecient (as a function of wind speed, Cp (v)) for the
Enercon E82 turbine.1
Figure 6: Power coecient function for Enercon E82 turbine
Surprisingly, some manufacturers, like Vestas and Siemens, do not show the Cp (v)
curve.
Importantly, the maximum value of 0.5 is very close to the Betz limit of
0.59. Hence, we conclude that wind turbine technology is mature, with very
little room for eciency improvements. Any substantial improvement in the
fraction of energy that we are able to extract from wind will have to come
from a radically dierent technology.
3.3
3.3.1
If windmills are packed too close to each other, the upwind ones will cast
wind-shadows on the downstream ones. According to MacKay (2009), the
distance between turbines should be at least ve times their diameter (d), as
in gure 7. Andrews and Jelley (2007) suggest a higher spacing. For example,
they state that a spacing of 8d downwind by 5d crosswind decreases the farm
output by around 10% when compared to the output of the turbines sited
separately. We follow the more optimistic 5d by 5d assumption and ignore
losses.
Figure 7: Wind farm layout (source: MacKay (2009))
5d
3.3.2
At a constant speed, the power of wind is given by (4). Since speed varies
through time, we can compute the expected power of the wind as
1
Av 3 f (v)dv
E[Pwin ] =
2
10
where f (v) denotes the density of the wind speed distribution. For a turbine
with power coecient function Cp (v), the expected output power is
1
E[Pout ] = Cp (v) Av 3 f (v)dv
2
Dene the expected output power density as
E[Pden ] :=
E[Pout ]
Land area
(7)
Assuming that the turbines are spaced at 5 diameters (land area = (5d)2 )
and replacing A = d2 /4, we get
=
Cp (v)v 3 f (v)dv
(9)
200
Interestingly, the power density does not depend on the diameter (d) of the
rotor. The ds canceled because bigger windmills have to be spaced further
apart. This means that wider turbines will not give us more electricity
(though higher turbines will because the wind speed increases with height).
However, larger turbines typically have some advantages, like lower maintenance and grid infrastructure costs. Also, many ridges only allow a single
line of turbines, so wider rotors will produce more electricity.
To illustrate (9), we compute the expected output power density for the
Enercon E82 turbine. The Cp (v) function is given by the manufacturer for
speeds up to 25 m/s (see gure 6, though the turbine brochure gives the
actual numbers in a table). We assume that the wind speed at the site
follows the Rayleigh distribution in (5) with a mean of 6 m/s. We use the
standard air density of = 1.225 kg/m3 . Discretizing the integral in (9) over
the 25 wind-speed points given by the manufacturer (v = 1), we get
E82
E[Pden
]
25
1.225
Cp (v)v 3 f (v) = 3.15 W/m2
200
v=1
(10)
11
(11)
Table 2 shows an estimate of the power density for the sample of turbines in
table 1, assuming that each turbine occupies a square of land of size (5d)2 .
Excluding the biggest 7.5 MW turbine that seems to be an outlier (probably
needs a bigger area of land), the average rated power density for all turbines
is
rated
Pden
10 W/m2
Dene the load factor as the ratio of actual generated power over the
installed capacity:
output power
Load Factor :=
(12)
rated power
The expected output power density dened in (7) can also be written as
E[Pden ] :=
E[Pout ]
E[Pout ]
rated power
=
land area
rated power
land area
12
(13)
Turbine
Enercon E53
Vestas V90
Enercon E82
Siemens SWT 113
Vestas V112
Siemens SWT 120
Enercon E126
Rated
Power
(kW)
Rotor
diameter
(m)
800
000
000
300
075
600
500
52.9
90
82
113
112
120
127
2
2
2
3
3
7
Area per
turbine
(m2 )
69
202
168
319
313
360
403
960
500
100
225
600
000
225
Rated power
density
(W/m2 )
11.4
9.9
11.9
7.2
9.8
10.0
18.6
The advantage of this approach is that we can use the load factor actually
observed in a given region. This will capture the interaction between the
actual varying wind speed and the turbines eciencies over all wind farms
installed in that region. Hopefully, this will be a reasonable estimate for
other future wind farms. The following section describes the load factor in
Portugal.
4.1
Load factor
To compute the load factor dened in (12), we use the hourly wind production for all wind parks in Portugal during 2010 (data available from REN).
According to DGEG, the installed wind capacity at the beginning of 2010
was 3507 MW and at the end was 3865 MW. We assume that the capacity
increased linearly throughout the year. This results in a time series with
8760 hourly load factors.
Figure 8 shows the histogram of the load factor.2 Across the full year of
2
Note that since our data series is the aggregate output of all wind parks in Portugal,
we never observe a load factor of 1, even though it is possible that individual parks may
operate at full capacity for some hours.
13
600
Number of hours
500
400
300
200
100
0.1
0.2
0.3
0.4
0.5
Load factor
0.6
0.7
0.8
0.9
2010, the mean load factor was 0.28 and the median was 0.22. Hence, we use
the following reference value:
E[Load Factor] 0.28
(14)
Note that this may be a somewhat optimistic estimate because the existing
wind parks very likely already took many of the best locations.
For example, if we apply this average load factor to the Enercon E82
turbine, we get an expected output power density
rated
E82
= 0.28 11.9 = 3.33 W/m2
] = E[Load Factor] Pden
E[Pden
(15)
14
(16)
This will be the main model to estimate the wind energy potential in Portugal.
4.2
Wind intermittency
Figure 8 already shows a lot of variability in the load factor, which is a direct
result of wind speed variability through time. In fact, wind speed intermittency is commonly pointed out as the main challenge to the integration of
signicant wind production in a national electricity grid. Hence, we take a
brief detour to examine the distribution of the load factor in more detail.
Figure 9 shows the distribution of the load factor for all wind parks in
Portugal during 2010, conditional on the the season of the year and the time
of day. First, we note that the load factor depends on the season of the year.
Winter shows the highest values with a median load around 0.4, while Summer shows the lowest values with median loads always below 0.2. The higher
availability in Winter is good as it matches the higher demand for heating.
Second, the load varies throughout the day: the median load factor decreases
in the early afternoon in all seasons, except Winter. This is unfortunate as
these are hours of high demand. If the standards of living increase and air
conditioning becomes more common in Portugal, the demand may peak in
the middle of hot summer days, exactly when wind is less available. Finally,
we note that the distribution of the load factor is always very wide. There is
a consistently large range between the 25th and 75h percentiles and the tails
of the distribution extend a lot beyond those points, reaching values very
close to zero in almost every hour of every season (Summer is an exception,
but because the distribution is packed in lower values).
In summary, wind energy is very intermittent. Currently, since wind has
dispatch priority in Portugal, the intermittency of wind is compensated by
natural gas power plants, which have fast online/oine times. In the future,
the intermittency of wind may also be compensated by better electricity
storage techniques (for example, batteries in electric vehicles), by a super
grid able to integrate many countries, or by other renewable technologies.
In particular, solar energy is probably a good match to wind: solar will be
stronger exactly during the periods where wind is weaker (the middle of the
day and the Summer months).
15
Spring
0.8
0.8
Load factor
0.6
0.4
0.6
0.4
0.2
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
0.2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Load factor
Winter
1
Hour of day
Hour of day
Autumn
0.8
0.8
Load factor
0.6
0.4
0.6
0.4
0.2
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
0.2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Load factor
Summer
1
Hour of day
Hour of day
16
Table 3 shows the energy demand in Portugal during 2010 (source DGEG).
The amount of electricity consumed was near 50 TWh and the total primary
energy was 206 TWh.
Table 3: Annual energy demand in Portugal, 2010.
(Source: DGEG)
Demand
1
1.1
1.2
1.3
2
3=1+2
TWh
Total Electricity
49.90
Electr. from Wind
9.18
Electr. from other Renewables
16.95
Electr. from fossil fuels and imports
23.77
Primary Energy for other usages 156.29
Total Energy Demand
206.18
To estimate how much energy we can generate from wind, we use equation
(16) for dierent areas of Portugal covered by windmills. More precisely, we
nd the required area for wind to supply each of the demand items described
in the previous table.
Table 4 shows the results. We estimate that we need 374 km2 of land to
generate the 9.18 TWh of wind energy in 2010. According to DGEG, there
were 208 wind farms in Portugal, so it seems that our model is reasonably
well calibrated (it would be interesting to know the actual area of the existing
wind farms, but this information is not available).
With 2.2% of the area of Portugal, we would be able to generate all of
the electricity demanded in 2010. This would require less than 5 times the
currently installed wind hardware. With 9.1% of the area of Portugal, we
would be able to generate enough wind energy to cover all our primary energy
needs from all sources!
17
Land area
(km2 )
(%)
Expected
total
output
power
(MW)
2 034
374
691
969
2.2%
0.4%
0.7%
1.0%
5
1
1
2
6 372
6.9%
17 841
8 406
92 300
9.1%
100.0%
23 537
258 440
696
048
935
713
Total Electricity
Electr. from Wind
Electr. from other Renewables
Electr. from fossil fuels and imports
156.29 Primary Energy for other usages
206.18 Total Energy Demand
2 263.93 Total area of Portugal
References
Andrews, J., and N. Jelley, 2007, Energy Science principles, technologies,
and impacts. Oxford University Press.
Esteves, T. M. V., 2004, Base de dados do potencial energetico do vento em
Portugal, Master Thesis, Universidade de Lisboa.
Gore, A., 2009, Our choice - a plan to solve the climate crisis. Bloomsbury.
MacKay, D. J., 2009, Sustainable Energy without the hot air. UIT Cambridge Ltd.
18