Wind Speed Probability Distribution and Parameter Estimation Methods

Abstract

Abstract A proper wind speed probability distribution model and an appropriate parameter estimation method can greatly improve the accuracy of prediction by indicators of wind energy. Therefore, in this paper, four groups of wind speed data were collected over a year from different altitudes in 2 typical coastal areas of Zhejiang Province, and then fitted by five wind speed probability distribution models, including Weibull, Rayleigh, Gamma, logarithmic normal and Inverse Gaussian distributions. After verification by multiple goodness-of-fit indicators, it was found that Weibull distribution was the optimal model for fitting the wind speed in the studied regions. The maximum likelihood estimation (MLE) method, power density (PD) method, moment method (MM), empirical method of Justus (EMJ), alternative maximum likelihood method (AMLM) and least-squares method (GM) were then used to calculate Weibull parameters at the four different places and for fitting accuracy analysis. The results showed that MLE, PD and MM methods achieved highly similar results in each of the goodness-of-fit indicators. Moreover, the Weibull distribution obtained by these three methods accorded better with the actual wind speed distribution, compared with that obtained by the other 3 methods.