Statistical Analysis of Hourly Wind Speed Data from Some Burundian Stations Using Beta Probability Density Functions

M. Bashahu, and M. Buseke

Department of Physics and Technology, Institute of Applied Pedagogy, University of Burundi, Burundi

Abstract: A 4-years period’s hourly wind speed (v) data from four stations in Burundi have been statistically analyzed in this paper. Attempts have been made to fit the monthly relative frequency distributions of those data with beta probability density functions (β-PDFs), which are quite different from Weibull PDFs commonly met in that kind of analysis. For that effect, the v quantity has been replaced by the relative wind speed (vr = v/vmax), the values of which range from 0 to 1. The β-PDFs fitting the observed monthly relative frequency distributions of the relative wind speed data have been implemented for the twelve (vr) data sets of each station. For every (vr) data set, the effectiveness of the fitting procedure has been checked through three complementary statistical tests, namely: the mean bias error (MBE), the root mean square error (RMSE) and the t-statistics. For any of the 48 (vr) data sets, the MBE and the RMSE have been found equal to or close to zero, and t was much lower than its critical value (tc = 3.250 for a confidence level γ = 99.5% and a number of degrees of freedom n-1 = 9). All the theoretical formulations (β-PDFs) obtained so far are therefore very good fits of the relevant observed monthly relative frequency distributions of hourly vr data. Moreover, contrarily to the Weibull PDFs, those β-PDFs accurately represent the probabilities of observing zero or very low wind speed. Consequently, they should be very useful input data in projects of wind energy conversion and/or storage systems at the stations of this study.

Key words: β-PDFs, MBE, relative frequency distributions, RMSE, relative wind speed, t-statistics