Documente Academic
Documente Profesional
Documente Cultură
FrequencyofWindSpeed
1600
ni:Numberofoccurrencesofwi (hrs/yr)
Frrequencyni (hourss/year)
1400
1200
Module2:BasicWindSpeed 1000
Statistics 800
600
MohamedA.ElSharkawi 400
DepartmentofElectricalEngineering
University of Washington
UniversityofWashington 200
Seattle,WA98195
http://SmartEnergyLab.com 0
Email:elsharkawi@ee.washington.edu 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
ProbabilityDistributionofWindSpeed Probability(p)
k
p
0.2 n
0.18 pi i i 1
0 6
0.16 N i 1
Probability(pi)
0.14
0.12
Assumehourlysampling
0.1
0.08 ni : frequency of wind speed wi (number of hours/year)
0.06
0.04
N : total number of hours in one year 8760 hrs
0.02 pi : probability of wind speed wi
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
k : Number of wind speed values
WindSpeedwi (m/s)
Mohamed A. El-Sharkawi, University of Washington 2.3 Mohamed A. El-Sharkawi, University of Washington 2.4
1
2/22/2009
AverageWindSpeed VarianceandStandardDeviation
1 1 k 2
ni wi wave
k
2
wave
N
n w
i 1
i i
Variance
N i 1
1 k 2
wi : i th wind speed Standard Deviation ni wi wave
N i 1
k : total number of wind speed values
Variance andstandarddeviationaremeasures
ofthestatisticaldispersionofthedata
Mohamed A. El-Sharkawi, University of Washington 2.5 Mohamed A. El-Sharkawi, University of Washington 2.6
CumulativeDistributionFunction ProbabilityDensityFunction(pdf)
b
Fab pwa wi wb pi Itisconvenienttomodelthediscrete wind
i a
speed probability data p(w) by a continuous
speedprobabilitydatap(w)byacontinuous
Fab : the sum of all probabilities betwwen wa and wb mathematicalfunctionf(w)
Note that Thisisparticularlyimportantifthesample
obability
F pi 1 widthisuneven
i 1
pro
pw f w
a b WindSpeed(w)
Mohamed A. El-Sharkawi, University of Washington 2.7 Mohamed A. El-Sharkawi, University of Washington 2.8
2
2/22/2009
ProbabilityDensityFunction(pdf) MainStatisticsofpdf
Theareaunderthepdf isunity
f(w)ProbabilityDensity f w dw 1
obabilityandpdf
0
Function
Thecumulative distributionfunction
f(w)
f(w)
Fab f w dw
b
Pro
pdff
a
Windspeed(w) a b w
Mohamed A. El-Sharkawi, University of Washington 2.9 Mohamed A. El-Sharkawi, University of Washington 2.10
MainPropertiesofpdf DistributionFunctionsforWindSpeed
Severaldensityfunctionscandescribewind
Average wind speed wave w f ( w) speed
d
0
Weibull distributionfunction(Twoparameter
distribution)
Wind Variance 2 w wave f ( w)
2
Rayleigh distributionfunction(Oneparameter
0 distribution)
Normal distributionfunction(symmetrical
Wind Stndard Deviation w w
2
ave f ( w) function)
0
Mohamed A. El-Sharkawi, University of Washington 2.11 Mohamed A. El-Sharkawi, University of Washington 2.12
3
2/22/2009
Wiebull DistributionFunction
k 1
k w
f w e w c
k
cc
Wiebull Distribution
k 0; c 0
Wiebull distributionparameters
c:Scaleparameter
k:Shapeparameter
Mohamed A. El-Sharkawi, University of Washington 2.13 Mohamed A. El-Sharkawi, University of Washington 2.14
ConstantScaleParameter(c) ConstantShapeParameter(k)
0.9
0.8
1 c=1 k=2
k=2.7
ebull distribution function
c=1 07
0.7
0.3 c=3
0.4
02
0.2
Wie
0.2 0.1
0
0 0 0.5 1 1.5 2 2.5 3 3.5 4
0 0.5 1 1.5 2 2.5 3 3.5 4 Wind Speed (m/s)
Wind Speed (m/s)
Mohamed A. El-Sharkawi, University of Washington 2.15 Mohamed A. El-Sharkawi, University of Washington 2.16
4
2/22/2009
AverageWindSpeed GammaFunction
k
w wave c x1/ k e x dx
wave c x1/ k e x dx; where x
0 c 0
Gamma function ()
Theequationfortheaveragewindspeedis
similartotheGammaFunction y x y 1 e x dx
0
NumericalvaluesforthegammafunctionExist
Mohamed A. El-Sharkawi, University of Washington 2.17 Mohamed A. El-Sharkawi, University of Washington 2.18
GammaFunction GammaFunction(1+1/k)
(*)availableintableformsand
(y)
y x
canbecomputedbyseveral
wave c x 1/ k x
e dx y 1 x
e dx mathematicalsoftwarepackages
p g
0 0
k y=(1+1/k) (y)
(*)isthegammafunctionat* 0.5 3 2
1 2 1
1.5 1.667 0.903 y
Ify=1+1/k;theaveragewindspeedisthe 2 1.5 0.886
weighted
h d gammafunction
f 2.5 1.4 0.887
wave c 1 1 / k
3 1.333 0.8929
3.5 1.2857 0.8997
4 1.25 0.906
Mohamed A. El-Sharkawi, University of Washington 2.19 Mohamed A. El-Sharkawi, University of Washington 2.20
5
2/22/2009
ProbabilityDensityFunction Example
w
a
k
Wiebull functionrepresentingwindspeedata
pw wa f w dw
d e c given site has c=5 and k=1 2 Compute the
givensitehasc=5andk=1.2.Computethe
wa numberofhoursperyearwhenthewind
speedis4m/sorgreater.
k k
wb w w w
k
4
1.2
a b a
pwa w wb f w dw e
c
e c
p w 4 e c
e 5
0.485
wa
Mohamed A. El-Sharkawi, University of Washington 2.21 Mohamed A. El-Sharkawi, University of Washington 2.22
1 k
n w
0.8
0.6
wave i i
0.4
N i 0
bull
1 k 2
Wieb
02
0.2
00 ni wi wave
0.5 1 1.5 2 2.5
Wind Speed (m/s)
3 3.5 4
N i 1
Mohamed A. El-Sharkawi, University of Washington 2.23 Mohamed A. El-Sharkawi, University of Washington 2.24
6
2/22/2009
ComputationofWiebull Parameters;
ApproximateMethod
SecondStep:Kandc canbeapproximatedby
1.086
Rayleigh(chi2)Distribution
k ; for 1 k 10
ave
w
wave
c
1 1 / k
Mohamed A. El-Sharkawi, University of Washington 2.25 Mohamed A. El-Sharkawi, University of Washington 2.26
Rayleigh(chi2)DistributionFunction RayleighDistributionStatistics
yleigh distribution function
2
w
w
f w
4 wave 2
0.5
e wa
2
2 wave
pw wa e
0.4 4 wave
0.3
4 2
0.2
2 1 wave
Ray
0.1 w ave
7
2/22/2009
AdvantageofRayleighDistribution
1 k
wave ni wi
N i 0
Thevarianceisonlyafunctionoftheaverage NormalDistribution
windspeed.
Themeanwindspeediseasily computedatany
site
Rayleighdensityfunctionissimpler touse
Mohamed A. El-Sharkawi, University of Washington 2.29 Mohamed A. El-Sharkawi, University of Washington 2.30
ComparisonbetweentheDistribution
NormalDistributionFunction
0.4 functions
0.35
0.3 Weibull
b ll isthemostaccurate
h
Normal distribution function
Normaldistributionistheleastaccurate
0.25
2
NormalandRayleighareeasiertouse
0.2
w wave 2
1 However,anyofthefunctionsmaybebetter
0 15
0.15 f w e 2 2
i
inaparticularrangeofwindspeeds
ti l f i d d
N
2
0.1
w ave
0.05
0 0.5 1 1.5 2 2.5 3 3.5 4
Wind Speed (m/s)
Mohamed A. El-Sharkawi, University of Washington 2.31 Mohamed A. El-Sharkawi, University of Washington 2.32
8
2/22/2009
CorrelationofWindSpeeds
Windspeedinonesitecouldcorrelatewith
d d ld l h
CorrelationandAutocorrelation thatinothersites
Windspeedpatternscouldexhibit
repeatability
Correlationindicatesthestrength
Correlation indicates the strength and
and
directionoflinear relationshipbetween
two randomvariables
Mohamed A. El-Sharkawi, University of Washington 2.33 Mohamed A. El-Sharkawi, University of Washington 2.34
CorrelationCoefficient Autocorrelation
Ameasureofthecorrelationbetweentwo random Amathematicalrepresentationofthedegree
variables xandy
x and y of similarity between a given time series and a
ofsimilaritybetweenagiventimeseriesanda
xy
E x x y y laggedversionofitselfoversuccessivetime
intervals
x y Isamathematicaltoolforfindingrepeating
patterns
xy:CorrelationCoefficient
aperiodicsignalburiedundernoise
i di i l b i d d i
X:Expectedvalues identifyingthecomponentfrequencyinasignal
: standarddeviations crosscorrelation ofasignalwithitself
Mohamed A. El-Sharkawi, University of Washington 2.35 Mohamed A. El-Sharkawi, University of Washington 2.36
9
2/22/2009
AutocorrelationofSingleSignal AutocorrelationofdiscreteSignals
y Y y
nk
i ik Y
t1 tk tn Rk i 1
y Y
n
2
i
i 1
R : Autocorrelation
k : Sampling lag time
SlidingWindow
yi : Sample at time i
Y : Average of all samples
Mohamed A. El-Sharkawi, University of Washington 2.37 Mohamed A. El-Sharkawi, University of Washington 2.38
AutocorrelationofContinuous
AutocorrelationofContinuousSignals
Function
E yi yi k
Rk
2 Rk f (t ) f (t k ) dt
E : Expected Value
yi : Sample
p at time i f : Function to be evaluated
: Mean value
: Variance
Mohamed A. El-Sharkawi, University of Washington 2.39 Mohamed A. El-Sharkawi, University of Washington 2.40
10