Evaluation of Probability Distribution Functions For Wind Speed Analysis: A Case Study of Addis Ababa
No Thumbnail Available
Date
2020-09
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Selection of best fit probability distribution model to the wind data sets is very important for
reducing uncertainties in the extreme wind speed modeling at a given site. This study presents
extreme wind speed analyses by using two approaches: The Block Maxima and Peaks Over
Threshold, and compare of the results of both methods for the data from Addis Ababa; Bole
wind speed recording station. Wind data collected include wind speed and wind direction,
recorded for 63 years (1954-2016) with three-hours’ time interval. Seven Two-parameter, five
Three-parameter, four Four-parameter, and Five-parameter Wakeby distributions are fitted to
the Block Maximum data series, and Generalized Pareto distribution is fitted to the Peaks
Over Threshold series. Three parameter estimation techniques were considered for estimating
parameters involved with these distributions namely, Maximum Likelihood, L-moments, and
Methods of Moments. The best fit models to the data are selected by examining ProbabilityProbability
Plots
and
four
goodness-of-fit
statistics:
Root
Mean
Square
Error,
Coefficient
of
Determination,
Kolmogorov-Smirnov,
and
Cramer-VonMises,
at
95
percent
confidence
level.
The
L-Moments estimation method has performed better for calculating the parameters of
most of the distributions while the Method of Moments is the preferred method for obtaining
the parameters of the JohnsonSB and Kumaraswamy distributions. The results showed that
the JohnsonSB distribution gives a best fit to the block maximum series. The Peaks Over
Threshold method with 2 peaks per year gave better results than the Block Maximum method;
as a result, Peaks Over Threshold method is recommended for design. Wind direction analysis
along with a wind rose chart for the study area is also provided. Analysis showed that most of
the winds come from the East and East-Southeast direction with the maximum magnitude of
3.60- 5.70 m/s. Finally, the selected distribution model is used for forecasting the extreme
wind speeds for return periods of 5, 10, 20, 50 and 100 years.
Description
Keywords
Block Maximum, Extreme Wind Speed, Goodness-of-Fit, L-Moments, Maximum Likelihood, Peaks Over Threshold, Probability Distribution Model