Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
Comparative analysis of five probability density functions applied to wind speed
Submission Author:
Darío Gerardo Fantini , DF
Co-Authors:
Darío Gerardo Fantini, Mario Benjamim Baptista de Siqueira, Reginaldo Nunes da Silva, Antonio Brasil Junior, Rafael Castilho Faria Mendes
Presenter: Mario Benjamim Baptista de Siqueira
doi://10.26678/ABCM.COBEM2023.COB2023-1823
Abstract
For feasibility analysis of a wind power plant knowing the probability density function (PDF) of wind speed, as accurately as possible, of the geographical location involved is relevant for a reliable estimate of energy generation in different months or time intervals in the year. Usually, the PDF is determined for the site of interest that enables the choice of the type of turbine and, therefore, the energy potential of the site, using the power curve or the power coefficient of the turbine. In most cases, the two-parameter Weibull distribution is the chosen function, but the Weibull distribution does not estimate accurately, especially in the cases with a bimodal distribution, because these other PDFs are proposed. Thus, in this work chosen, based on the literature, the PDFs Weibull of two parameters, Weibull of two components, Gamma, Log-Normal and Log-Logistic. For this work were used the wind speed and direction data at 50m height was acquired from NASA Power, with the hourly average for the sites of Aparecida de Goiânia and Itumbiara in Goiás state, and Fortim in Ceará state. The time series comprises the years 2002 to 2021. To obtain more relevant indicators, which allow inferring which of the PDFs is the best fit for the distribution of the actual wind speed data, the following procedures were implemented. From the complete temporal series (2002-2021) the samples are grouped by month, from each new grouping the different PDFs of each month are obtained, which are used to determine the energy generated by the chosen turbine (Eturb). Then the original series is separated into samples (new time series) representative of each month of each year X(year, month), which are used to determine the Eturb(year, month) respectively. As a function of the values of Eturb(year, month) and Eturb for each PDF the RMSE, the MAE and the MAPE are determined. Finally, it is determined that the two-component Weibull is the best fit for the wind speed distribution in the different analyses (hourly, monthly and annual), giving a more reliable visibility of the real frequency of wind speed distribution, but this significant visual improvement is not reflected in the calculation of the average power density available, average energy and when compared to a real sample, compared to the Weibull PDF. When compared to the other PDFs analysed it is concluded that both Weibull PDFs are significantly more representative to describe the wind speed behaviour.
Keywords
Wind power, two-component Weibull, Comparative probability density functions, Two-component mixture Weibull, Mixture Weibull, Wind Power Density, wind energy, Brazil's wind resources
