Modelling of Wind Speed Data Using Weibull, Champernowne & Johnson SB Distributions

Abstract

The wind speed data must be modelled correctly using the appropriate
probability distributions so that we can assess the wind energy potential in
an effective way. In this paper, wind speed data from three different sites
are modelled using Weibull, Champernowne & Johnson SB distributions.
Their parameters are estimated using Maximum Likelihood Estimation
(MLE) and Least Square Method. The fit is validated using the Goodness
of Fit tests - Kolmogorov-Smirnov (KS) test, Akaike Information Criterion
(AIC) and Bayesian Information Criterion (BIC). The analysis shows that
for the first site, Johnson SB gives the best fit with lower BIC. For second
and third sites, Champernowne distribution gives the best fit. The performance
of Weibull distribution was not good, as it scored high AIC and BIC
values across all sites. From the study, we are able to suggest Johnson SB
and Champernowne distributions as an alternative to conventional Weibull
distribution. It is essential to consider the local wind characteristics while
selecting distributions for wind energy modelling