Wind Energy Systems MASE 5705 Spring 2014, L3+ L4L4.pdf(Random Data, Analysis and Measurement...
Transcript of Wind Energy Systems MASE 5705 Spring 2014, L3+ L4L4.pdf(Random Data, Analysis and Measurement...
1
Wind Energy Systems MASE 5705
Spring 2014, L3+ L4
Review of L1 + L2
2
3
(2.75) p. 63 , Second Edition
4
(2.74. p/ 63, 2nd edition
Dynamic
Power
5
Wind power
Grid
Mechanical – Electrical Conversion Chain Efficiency
(based on “Wind Turbines” Erich Hau, Springer, 2006)
6
Site potential for
Wind Energy Production
We have the database --- a large quantity
of data on wind speed and its direction.
For example, 8760 hourly measurements
for one year.
We wish to calculate the wind-energy
production potential.
7
Two Widely Used Methods
1. Tabular approach or Method of Bins
2. Closed-form expressions for wind-
speed probability distribution
functions.
8
1. Elements of probability and statistics
2. Method of bins (Bin means wind-speed
interval; typically this interval is held
constant. Each bin contains data for that
interval)
3. Weibull and Raleigh distribution functions.
9
1. As for the coverage of probability and
statistics, we follow the text; the
emphasis is on applications and not on
precision.
2. In the beginning, the concepts of
probability may seem abstract. With the
solution of problems, our understanding
and appreciation of these concepts and
their utility get much better.
10
For the present assume:
11
Given N independent measurements of wind
speed: Ui, i= 1 , …, N
Mean wind speed = average wind speed = expected wind speed
12
……..……..(2.39)
p.54
(A)
(B)
13
……..……..(2.40)
p.54
However, Eq. (A) is considered a
‘better’ estimator.
(Random Data, Analysis and
Measurement Procedures, Bendat, and
Piersol, Wiley 3rd Ed., 2000, p. 87.)
Here after, only Eq. (A) will be used.
14
Notational Differences (2.20) p. 40
15
Total mean Turbulent
or random
16
17
Wind-speed data are arranged into bins
or wind-speed intervals, which typically
span 0.5 m/s. ( These procedures are
often governed by specifications from
ASME, AWEA and such organizations.)
18
The data were continuously measured every hour for one year ≡ 8760
measurements. The site refers to MOD-2 (Boeing) 91.4 m dia turbine; cut-
in speed=6.25 m/s and cut-out speed=22.5 m/s.)
19 Bin ≡ wind velocity interval, specifications often require 0.5 m/s bins.
A Sample of Database Arranged into Bins
BIN Duration
Min(m/s) Max(m/s) Δ t k (h/y)
1 0 6.25 2147
2 6.25 6.75 416
3 6.75 7.25 440
4 7.25 7.75 458
5 7.75 8.25 468
6 8.25 8.75 470
7 8.75 9.25 466
8 9.25 9.75 453
9 9.75 10.25 435
10 10.25 10.75 410
11 10.75 11.25 381
12 11.25 11.75 349
13 11.75 12.25 314
14 12.25 12.75 278
15 12.75 13.25 242
16 13.25 13.75 208
17 13.75 14.25 175
18* 14.25 >22.4 648
19 >22.4 2
Totals: 8760 = N
Wind Speed
[from: Wind turbine
Technology, David A.
Spera, editor , 1994,
p.222, ASME publication]
*For Bin 18,
The reason for the much larger
interval (22.5-14.25 = 8.5 m/s)
is not known, we will revisit the
data base.
20
Similarly
21
(A)
(C)
N–1 j=1 N j=1 (2.47) p. 55
A descriptive account with illustrations (for mathematical details see Bendat and Piersol, ibid.)
1a) Probability of observing Ui or p(Ui) for the discrete values of Ui
1b) (corresponding) cumulative distribution function F(Ui)
2a) Probability density function f(U) for the continuous values of U.
2b) (corresponding) cumulative distribution function F(U)
22
Experiment:
23
24
We also need
25
An illustration:
wind speed histogram
26
i U i m i f(u i ) F(U i )
1 0 0 0 0
2 1 0 0 0
3 2 15 0.071 0.071
4 3 42 0.199 0.270
5 4 76 0.360 0.630
6 5 51 0.242 0.872
7 6 27 0.128 1.000
N= 211
(ref: Wind Energy Systems, G. L. Johnson, Prentice Hill, NJ, 1985, p. 53.)
27
Formally,
Wind speed variations represent continuous
random functions, that is, continuous
mathematical functions. Bypassing
mathematical details we accept the next set
of definitions:
28
Cumulative distribution function
29
(2.56, p.58)
30
U
f(U)
Probability Density Function f(U)
U
F(U)
Cumulative Distribution Function F(U)
1
0
0
31
The probability density function f(U) has the following fundamental properties
See page 58, 2nd Ed.
Wind Power Classes (for later reference ), p. 67
(Class 4 or greater are suitable for electrical utility applications)
Hub
Height
10 m (33 ft) 30 m (98 ft) 50 m (164 ft)
Wind
Power
class
Power
Density
W/m2
Speed m/s
(mph)
Power
Density W/m2 Speed m/s (mph)
Power
Density W/m2
Speed m/s
(mph)
1 0-100 0-4.4 (0-9.8) 0-160 0-5.1 (0-11.4) 0-200 0-5.6 (0-12.5)
4.4-5.1 5.1-5.8 200-300 5.6-6.4
2 100-150 (9.8-11.5) 160-240 (11.4-13.2) (12 5-14.3)
5.1-5.6 5.8-6.5 300-400 6.4-7.0
3 150-200 (11.5-12.5) 240-320 (13.2-14.6) (14.3-15.7)
5.6-6.0 6.5-7.0 400-500 7.0-7.5
4 200-250 (12.5-13.4) 320-400 (14.6-15.7) (15.7-16.8)
6.0-6.4 7.0-7.4 500-600 7.5-8.0
5 250-300 (13.4-14.3) 400-480 (15.7-16.6) (16.8-17.8)
6.4-7.0 7.4-8.2 600-800 8.0-8.8
6 300-400 (14.3-15.7) 480-640 (16.6-18.3) (17.8-19.7)
7.0-9.4 8.2-11.0 800-2000 8.8-11.9
7 400-1000 (15.7-21.1) 640-1600 (18.3-24.7) (19.7-26.6)
32
Wind Data Distributed by Bins for Hypothetical
Site (average wind speed 5.57 m/s or 12.2 mph)
33
* fi = observed bin probability or probability density function for the i-th bin
(fi is also referred to as frequency)
BIN hi fi (0bserved)
Ui
Min(m/s) Max(m/s) hours/year (pct) Bin Avg.
Speed Σfiui
hi/8760 (m/s) 1 0 0 80 0.91% 0 0.0000 2 0 1 204 2.33% 0.5 0.0116
3 1 2 496 5.66% 1.5 0.0849 4 2 3 806 9.20% 2.5 0.2300 5 3 4 1211 13.82% 3.5 0.4838
6 4 5 1254 14.32% 4.5 0.6442 7 5 6 1246 14.22% 5.5 0.7823 8 6 7 1027 11.72% 6.5 0.7620 9 7 8 709 8.09% 7.5 0.6070
10 8 9 549 6.27% 8.5 0.5327 11 9 10 443 5.06% 9.5 0.4804 12 10 11 328 3.74% 10.5 0.3932
13 11 12 221 2.52% 11.5 0.2901 14 12 13 124 1.42% 12.5 0.1769 15 13 14 60 0.68% 13.5 0.0925
16 14 No upper
bond 2 0.02% -
N=8760 5.572
Wind Speed
Ū=
34
Bin
Pro
babili
ty f
(Ui)
Use of the closed–form
expressions for probability
distribution functions of wind
speeds.
35
36
37
38
By definition (2.60)(2.61) p. 59, 2nd Ed.
39
40
Weibull
Observed wind speed distributions for three New England locations and
the Weibull with the parameters k = 1.8 and c = 11. 5 mph
(From: Atmospheric Turbulence, H. Panofsky and J. Dutton, Wiley, 1984)
41
Wind speed U m/s
Approximation of the measured wind distribution on the island of Sylt by Weibull
functions. (From: Wind Turbines, by Erich Hau, Springer, 2006)
Pro
babili
ty D
ensity f
unction
Pro
babili
ty D
istr
ibution f
unction
We accept: (2.53) p. 58, 2nd Ed.
42
43
(2.54) p. 58, 2nd Ed.
An example
44
45
Repeat
46
47
48
49
Weibull Distribution Calculations (B2.6, p.618)
From an analysis of wind speed data (hourly interval
average, taken over a one year period), the Weibull
parameters are determined to be c = 6 m/s and k = 1.8 .
a) What is the average velocity at this site?
b) Estimate the number of hours per year that the wind
speed will be between 5.5 and 7.5 m/s during the year.
c) Estimate the number of hours per year that the wind
speed is above 16 m/s.
50
Pr. B (2.6) p 618, (pp. 58-59)
51
52
(2.61) and (2.62) pp. 59-60
53
54
55
56
57
58
Revisiting Weibull (2.60) (2.61), p. 59 2nd Ed.
59
60
k and c
61
62
63
64
65
Cf(
u)
For c=1
U/C = 0.7 = U
Cf(U) = 0.858 = f(U)
for c=10
U = 7 m/s
f(U) = 0.0858
Finding Weibull Parameters k and c
1. Least squares fit for the observed values
of bin frequency fi (e.g. subroutine:
lsqcurvefit), (not in the text.)
2. Analytical–Empirical Approach
(pp. 60-61)
3. Graphical Approach (p.61)
4. Closed-form approach
66
Although, not included in the text, the least-
squares-fit approach is very powerful and
perhaps the best.
We will first apply this approach and then
take up the other three. ( The database
refers to the earlier-treated hypothetical
site.)
67
68
BIN hours/year 0bserved bin Bin Avg. Speed (Ui) Σfiui Fi
No. Min(m/s) Max(m/s) frequency fi (m/s) Cumulative Frequency
1 0 0 80 0.91% 0 0.0000 0.009
2 0 1 204 2.33% 0.5 0.0116 0.032
3 1 2 496 5.66% 1.5 0.0849 0.089
4 2 3 806 9.20% 2.5 0.2300 0.181
5 3 4 1211 13.82% 3.5 0.4838 0.319
6 4 5 1254 14.32% 4.5 0.6442 0.462
7 5 6 1246 14.22% 5.5 0.7823 0.605
8 6 7 1027 11.72% 6.5 0.7620 0.722
9 7 8 709 8.09% 7.5 0.6070 0.803
10 8 9 549 6.27% 8.5 0.5327 0.866
11 9 10 443 5.06% 9.5 0.4804 0.916
12 10 11 328 3.74% 10.5 0.3932 0.954
13 11 12 221 2.52% 11.5 0.2901 0.979
14 12 13 124 1.42% 12.5 0.1769 0.993
15 13 14 60 0.68% 13.5 0.0925 1.000
16 14 14> 2 0.02% 14 0.000 1.000Total no. of
hours or
measurements
N= 8760
5.571804 1.000
Wind Speed
Least-Squares Approach
69
(2.66) p. 60 2nd Ed.
70
71
Analytical-Empirical Approach
pp. 59-60, 2nd Edition
72
Empirical-Analytical Method
73
Hypothetical Site
Parameters k c
least squares 2.117 6.293
empirical-analytical 2.066 6.291
74
The next graph shows how Weibull fi from these two methods and
also from Rayleigh (k=2) compare with the data or observed values
75
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 160
0.05
0.1
0.15
BIN Number
BIN
Fre
qu
en
cy
data
Weibull(lsq)
Weibull(empirical)
Homework
For the hypothetical site, study:
a.) the graphical method to find k and c.
b.) use closed-form method with Γ(n):
(σU/U)2 = Γ(1+2/k)/Γ2(1+1/k) – 1
c = U/Γ(1+1/k)
Hint: plot function and see where = (σU/U)2.
76
Homework
77
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 30.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
K
f(k)
Graphical Method (Hints)
78
y = a x + b Finally,
k = a and c = exp (-b/k)
Recommended Reading Stevens MJM and Smulders PT (1979)
“The estimation of parameters of the Weibull speed
distribution for wind energy utilization purposes.” Wind
Engineering 3(2): 132-145.
(If you need a copy, contact Gaonkar or Peters.)
The paper discusses five methods of calculating k and c. This includes
empirical and graphical methods but not the least squares method.
The article is written in an easy-to-read and descriptive style, and it
address some other type of WE issues as well.
79
Please submit all the homeworks and
projects in two separate files.
HW file: Weekly on Tuesdays
Project file: April 26, 2012
80
81
82
Set or
Set
83
(2.62), p. 60, 2nd Edition
84 (2.64) p. 60 2nd Edition
85
86
Most probable or most frequent wind speed
and
Energy Pattern Factor (EPF)
and
The closed-form expression for the Weibull
Case
87
88
89
f(u)
0.854
c=1
Energy Pattern Factor
90
91
(2.69) p. 61 2nd Edition
92
Table 2.4 p. 61, 2nnd
Edition
Raleigh distribution
A special case of Weibull
Distribution with
shape factor k = 2
# 2.4.3.3 (pp. 58 – 61)
93
94
(2.61) p. 59, 2nd Edition
Rayleigh Distribution
95
(2.59) p. 59, 2nd Edition
(p.618)
B.2.7 Rayleigh Distribution Calculations
Analysis of time series data for a given site has yielded an
average velocity of 6 m/s. It is determined that a Rayleigh wind
speed distribution gives a good fit to the wind data.
a) Based on a Rayleigh wind speed distribution, estimate the
number of hours that the wind speed will be between 9.5 and
10.5 m/s during the year.
b) Using a Rayleigh wind speed distribution, estimate the
number of hours per year that the wind speed is equal to or
above 16 m/s.
96
Pr. 2.7. , P. 618
97
For Raleigh distribution
98
(2.59) p. 59, 2nd Edition
99
100
Application of Raleigh and
Weibull to a specific database
101
Consider the earlier-treated database
for “Hypothetical Site.”
Wind Data Distributed by Bins for Hypothetical
Site (average wind speed 5.57 m/s or 12.2 mph)
102
BIN hiFrequency
(0bserved)Ui
Min(m/s) Max(m/s) hours/year (pct)Bin Avg.
Speed Σpiui
hi/8760 (m/s)
1 0 0 80 0.91% 0 0.0000
2 0 1 204 2.33% 0.5 0.0116
3 1 2 496 5.66% 1.5 0.0849
4 2 3 806 9.20% 2.5 0.2300
5 3 4 1211 13.82% 3.5 0.4838
6 4 5 1254 14.32% 4.5 0.6442
7 5 6 1246 14.22% 5.5 0.7823
8 6 7 1027 11.72% 6.5 0.7620
9 7 8 709 8.09% 7.5 0.6070
10 8 9 549 6.27% 8.5 0.5327
11 9 10 443 5.06% 9.5 0.4804
12 10 11 328 3.74% 10.5 0.3932
13 11 12 221 2.52% 11.5 0.2901
14 12 13 124 1.42% 12.5 0.1769
15 13 14 60 0.68% 13.5 0.0925
16 14No upper
bond2 0.02% -
8760 5.572
Wind Speed
103
= 0.136
104
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 160
0.05
0.1
0.15
BIN Number
BIN
Fre
qu
en
cy
data
Rayleigh
105
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 160
0.05
0.1
0.15
BIN
BIN
Fre
qu
en
cy
data
Weibull(lsq)
Rayleigh
Weibull(empirical)
Wind-speed data points are notoriously
fickle. Probabilities calculated from Weibull
(more so with Rayleigh) exhibit considerable
scatter for some cases. We will take a look
at this next.
106
Limitations of Closed-form expressions
for wind-speed distributions.
107
Ui
f(U
i)
1. As expected Weibull is better than Rayleigh.
2. The Weibull approximation is not satisfactory.
(Wind Energy Systems, Ibid.)
108
f(U
i) (from: Wind Energy
Systems, Ibid.)
The data shows double-peaked behavior The approximation is not
satisfactory.
109
f(U
i)
(k=2.11
c=11.96)
(from: Wind Energy
Systems, Ibid.)
Actually, wind-speed statistics always exhibit some scatter. Thus the above
approximation may be satisfactory.
Double-Peaked bi-Weibull Distribution
110
Conclusions 1. A generalized statement about the accuracy of
the Weibull or Rayleigh probability density
functions (pdf) is not possible. Always, Weibull is
more accurate than Rayleigh but the latter is
much simpler.
2. If we can find such a Weibull or Rayleigh pdf, the
selection of a specific WT or the prediction of its
power output becomes dramatically simplified.
3. It is always advantageous to explore the
feasibility of finding such closed-form
expressions.
111